<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Bernardo Caldas on Medium]]></title>
        <description><![CDATA[Stories by Bernardo Caldas on Medium]]></description>
        <link>https://medium.com/@bernardo-acaldas?source=rss-f1a3625e122c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*siQ9qu-RniPRgO8naRFzwQ.jpeg</url>
            <title>Stories by Bernardo Caldas on Medium</title>
            <link>https://medium.com/@bernardo-acaldas?source=rss-f1a3625e122c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 14:21:23 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@bernardo-acaldas/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Artificial Intelligence is Transforming Maintenance Applications in Industry]]></title>
            <link>https://bernardo-acaldas.medium.com/artificial-intelligence-is-transforming-maintenance-applications-in-industry-d67c9b448859?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/d67c9b448859</guid>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[sap]]></category>
            <category><![CDATA[industrial-maintenance]]></category>
            <category><![CDATA[fiori]]></category>
            <category><![CDATA[maintenance]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Thu, 14 Nov 2024 01:39:27 GMT</pubDate>
            <atom:updated>2025-01-23T17:35:22.557Z</atom:updated>
            <content:encoded><![CDATA[<p>How Artificial Intelligence is Transforming Maintenance Applications in Industry</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*lLs49V09eAIjhvKd" /><figcaption>Photo by <a href="https://unsplash.com/@mrsunburnt?utm_source=medium&amp;utm_medium=referral">Xu Haiwei</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Follow the link the complete post for free:</p><p><a href="https://bernardo-acaldas.medium.com/artificial-intelligence-is-transforming-maintenance-applications-in-industry-d67c9b448859?source=friends_link&amp;sk=ee6008dd017eb5a9bb25f2fb1952d84a">https://bernardo-acaldas.medium.com/artificial-intelligence-is-transforming-maintenance-applications-in-industry-d67c9b448859?source=friends_link&amp;sk=ee6008dd017eb5a9bb25f2fb1952d84a</a></p><p>Imagine a maintenance app that not only logs requests but also offers smart suggestions to streamline your operations. In today’s fast-paced tech landscape, artificial intelligence (AI) is becoming an essential tool for driving efficiency and innovation in business.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*jRRWfUXHDGCcGCB9ioQwTA.gif" /><figcaption>Final result</figcaption></figure><p>In this article, I dive into a practical example of how I created a SAP Fiori app focused on maintenance that integrates a generative AI model to provide intelligent recommendations. Whether you’re a developer, a maintenance professional, or a tech enthusiast, this demonstration offers valuable insights into how AI can enhance daily processes.</p><p><strong>The Power of SAP Fiori and OpenUI5 in Maintenance</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/373/1*xU673yNAYNjtOb9EIKk6ag.png" /></figure><p>SAP Fiori is a user experience (UX) framework that provides an intuitive interface for SAP applications, making business interactions simpler and more efficient. OpenUI5, an open-source JavaScript library, enables the development of responsive and feature-rich web applications.</p><p>In maintenance, creating and managing work orders is crucial to ensure seamless, effective operations. However, this process can be complex and prone to human error. Integrating AI into this workflow can change that reality.</p><p>How can we leverage AI to improve maintenance applications and assist professionals in their daily tasks?</p><p>I set out to develop a custom SAP Fiori app, focused on maintenance, that integrates a generative AI model. The goal is simple: when the user enters the title and description of a maintenance order, the app automatically generates a list of essential steps, helping to ensure that no critical task is overlooked.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*9d_M_eVvf0tmqbvkXHpieA.png" /></figure><p><strong>Explore the Demo: </strong><a href="https://fiori-app-form.vercel.app/"><strong>https://fiori-app-form.vercel.app/</strong></a></p><p><strong>Technologies Behind This Solution</strong></p><ul><li><strong>Frontend</strong>: Built with OpenUI5, providing a familiar and user-friendly interface for SAP Fiori users.</li><li><strong>Backend</strong>: A locally hosted API utilizing a large language model (LLM) to process user inputs and generate relevant suggestions.</li></ul><p><strong>How It Works</strong></p><ol><li><strong>User Input</strong>: The user enters the maintenance order title and description in the custom Fiori app.</li><li><strong>AI Processing</strong>: The data is sent to the API, which uses the AI model to analyze the input and generate a list of suggested steps.</li><li><strong>Suggestion Display</strong>: The app presents these suggestions, offering the user valuable guidance for completing the task.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*kFZ4GnRLW38XMISgDzZ-4Q.png" /></figure><h3>Demonstration in Action</h3><p>Here’s how the application works in practice:</p><ol><li><strong>Title</strong>: “Air Filter Replacement”</li><li><strong>Description</strong>: “The air filter of the compressor needs to be replaced due to particle buildup.”</li></ol><p><em>Upon entering this information, the user receives immediate suggestions:</em></p><ul><li>1. Turn off the compressor and disconnect it from the electrical power.</li><li>2. Locate the air filter, which is usually in a plastic or metal case.</li><li>3. Loosen the screws or clips securing the air filter box.</li><li>4. Remove the old air filter and inspect it for dirt and damage.</li><li>5. Insert the new air filter into the housing and secure it with the screws or latches.</li><li>6. Replace the air filter box on the compressor.</li><li>7. Turn on the compressor and check for air leaks.</li></ul><p><strong>Why This Matters</strong></p><p>This app demonstrates how AI can:</p><ul><li><strong>Optimize Processes</strong>: Automating essential steps, saving time and resources.</li><li><strong>Support Decision-Making</strong>: Providing insights that may not be immediately apparent.</li><li><strong>Improve Quality</strong>: Reducing errors and ensuring compliance with standard procedures.</li></ul><p><strong>Future Potential</strong></p><ul><li><strong>Advanced Customization</strong>: Tailoring suggestions based on historical data and company preferences.</li><li><strong>Scalability</strong>: Expanding this solution to other modules and processes beyond maintenance.</li><li><strong>Continuous Learning</strong>: Using user feedback to continually improve the AI model.</li></ul><p><strong>Try It Out</strong></p><p>Curious to see this in action? Check out the app here:</p><p><strong>Online Demo</strong>: <a href="https://fiori-app-form.vercel.app/">https://fiori-app-form.vercel.app/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d67c9b448859" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Revolucionando a Manutenção com um App Fiori que Gera Sugestões usando IA]]></title>
            <link>https://bernardo-acaldas.medium.com/revolucionando-a-manuten%C3%A7%C3%A3o-com-um-app-fiori-que-gera-sugest%C3%B5es-usando-ia-37724415601c?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/37724415601c</guid>
            <category><![CDATA[sap]]></category>
            <category><![CDATA[fiori]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[industria]]></category>
            <category><![CDATA[manutenção]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Mon, 11 Nov 2024 00:21:37 GMT</pubDate>
            <atom:updated>2025-01-24T19:56:02.757Z</atom:updated>
            <content:encoded><![CDATA[<p>Como a Inteligência Artificial está Transformando Aplicativos de Manutenção nas industrias.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*YqyK5Miy65N6yunO" /><figcaption>Photo by <a href="https://unsplash.com/@karanbhatia9?utm_source=medium&amp;utm_medium=referral">Karan Bhatia</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Follow the link to read for free: <a href="https://bernardo-acaldas.medium.com/revolucionando-a-manuten%C3%A7%C3%A3o-com-um-app-fiori-que-gera-sugest%C3%B5es-usando-ia-37724415601c?source=friends_link&amp;sk=7314da3707051e789c942e2356b1dd28">https://bernardo-acaldas.medium.com/revolucionando-a-manuten%C3%A7%C3%A3o-com-um-app-fiori-que-gera-sugest%C3%B5es-usando-ia-37724415601c?source=friends_link&amp;sk=7314da3707051e789c942e2356b1dd28</a></p><p><strong>Imagine um aplicativo de manutenção que não apenas registra solicitações, mas também oferece sugestões inteligentes para otimizar suas operações.</strong> No ritmo acelerado da tecnologia atual, a inteligência artificial (IA) está se tornando uma ferramenta essencial para impulsionar a eficiência e a inovação nos negócios.</p><p>Neste artigo, exploro uma implementação prática de como criei um <strong>aplicativo SAP Fiori voltado para manutenção</strong> que integra um modelo de <strong>IA generativa</strong> para fornecer sugestões inteligentes. Seja você um desenvolvedor, um profissional de manutenção ou um entusiasta de tecnologia, esta demonstração oferece insights valiosos sobre como a IA pode aprimorar processos cotidianos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/878/1*G08_NpkEt4XmOKK5cpNGfQ.jpeg" /></figure><p><strong>SAP Fiori</strong> é uma estrutura de experiência do usuário (UX) que oferece uma interface intuitiva para aplicativos SAP, tornando as interações empresariais mais simples e eficientes. O <strong>OpenUI5</strong> é uma biblioteca JavaScript de código aberto que permite o desenvolvimento de aplicativos web responsivos e ricos em recursos.</p><p>Na área de manutenção, a criação e o gerenciamento de <strong>ordens de manutenção</strong> são cruciais para garantir operações contínuas e eficazes. No entanto, esse processo pode ser complexo e suscetível a erros humanos. A integração da IA pode transformar essa realidade.</p><p><em>Como podemos utilizar a IA para melhorar os aplicativos de manutenção e auxiliar os profissionais em suas tarefas diárias?</em></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/275/1*v-prp26DO1b8KubY9BvyLQ.png" /></figure><p>Decidi desenvolver um <strong>aplicativo SAP Fiori personalizado</strong>, focado em manutenção, que integra um modelo de IA generativa. O objetivo é simples: quando o usuário insere o título e a descrição da ordem de manutenção, o aplicativo gera automaticamente sugestões de passos importantes, ajudando a garantir que nenhuma etapa crítica seja esquecida.</p><p><strong>Demo Online</strong>: <a href="https://fiori-app-form.vercel.app/">https://fiori-app-form.vercel.app/</a></p><p>Para construir essa solução inovadora, utilizei:</p><ul><li><strong>Frontend</strong>: Desenvolvido com <strong>OpenUI5</strong>, proporcionando uma interface familiar e amigável aos usuários do SAP Fiori.</li><li><strong>Backend</strong>: Uma API rodando localmente que utiliza um modelo de linguagem de grande porte (LLM) para processar as entradas do usuário e gerar sugestões relevantes.</li></ul><h3>Como funciona</h3><ol><li><strong>Entrada do Usuário</strong>: O usuário insere o <strong>título</strong> e a <strong>descrição</strong> da ordem de manutenção no aplicativo Fiori personalizado.</li><li><strong>Processamento da IA</strong>: Essas informações são enviadas para a API, que utiliza o modelo de IA para analisar os dados e gerar uma lista de sugestões de passos a serem seguidos.</li><li><strong>Exibição das Sugestões</strong>: As sugestões geradas são apresentadas no aplicativo, oferecendo ao usuário orientações valiosas para a execução da tarefa.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*QWk8QxvWFv9WXkmtljZd8w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Tqizun07N34Z5s2_nIfkyQ.png" /><figcaption>Sugestões para execução da tarefa</figcaption></figure><h3>Por que Isso é Importante</h3><p>A implementação deste aplicativo demonstra como a IA pode:</p><ul><li><strong>Otimizar Processos</strong>: Automatizando a geração de passos essenciais, economizando tempo.</li><li><strong>Apoiar Decisões</strong>: Fornecendo insights que podem não ser imediatamente aparentes.</li><li><strong>Melhorar a Qualidade</strong>: Reduzindo erros e garantindo a conformidade com procedimentos padrão.</li></ul><h3>Potencial Futuro</h3><p>As possibilidades futuras incluem:</p><ul><li><strong>Customização Avançada</strong>: Adaptar as sugestões com base em dados históricos e preferências da empresa.</li><li><strong>Escalabilidade</strong>: Implementar a solução em outros módulos e processos além da manutenção.</li><li><strong>Aprendizado Contínuo</strong>: Utilizar feedback dos usuários para aprimorar continuamente o modelo de IA.</li></ul><h3>Experimente Você Mesmo</h3><p>Curioso para ver a solução em ação? Acesse o aplicativo e teste:</p><ul><li><strong>Demo Online</strong>: <a href="https://fiori-app-form.vercel.app/">https://fiori-app-form.vercel.app/</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=37724415601c" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Elevating RCA and 5 why method to New Heights: The Fusion of AI and Root Cause Analysis]]></title>
            <link>https://medium.com/analytics-vidhya/elevating-rca-to-new-heights-the-fusion-of-ai-and-root-cause-analysis-6bd6811b87ce?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/6bd6811b87ce</guid>
            <category><![CDATA[rca]]></category>
            <category><![CDATA[root-cause-analysis]]></category>
            <category><![CDATA[analysis]]></category>
            <category><![CDATA[chatgpt]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Tue, 15 Aug 2023 19:58:57 GMT</pubDate>
            <atom:updated>2025-01-29T14:13:42.660Z</atom:updated>
            <content:encoded><![CDATA[<p>How AI is Transforming Problem Analysis in Businesses and Aiding in Finding More Effective Solutions</p><p>follow the free link: <a href="https://bernardo-acaldas.medium.com/elevating-rca-to-new-heights-the-fusion-of-ai-and-root-cause-analysis-6bd6811b87ce?source=friends_link&amp;sk=799b929ab87655d9a9db72bdb39d2534">https://bernardo-acaldas.medium.com/elevating-rca-to-new-heights-the-fusion-of-ai-and-root-cause-analysis-6bd6811b87ce?source=friends_link&amp;sk=799b929ab87655d9a9db72bdb39d2534</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SsHZiy87Cu0fqWjJISIBUA.png" /></figure><p>In the business world, tools like 5W2H and the 5 Whys are essential for unraveling and solving problems. While 5W2H provides a clear structure to understand challenges, the 5 Whys dive deep into the root cause of the issue. Together, they have guided teams in making informed decisions. But, in a digital age, how can we enhance these traditional approaches? Example of a form used nowadays.</p><p>AI Transforming the Analysis</p><p>With integrated artificial intelligence (AI), the traditional 5W2H and 5 Whys form is taken to another level. AI not only suggests answers but also optimizes them, making each analysis sharper and more effective. It’s the merging of the wisdom from classic tools with the dynamism of modern technology. Application in operation suggesting an action based on form responses.</p><p>Use Cases in Companies: Real Results for Modern Businesses</p><p><em>Process Improvement</em>: An automobile factory utilizes the AI-enhanced 5W2H and 5 Whys form to identify bottlenecks in the production line. With intelligent suggestions, the company manages to increase efficiency by 20% in just a few weeks.</p><p><em>Solving Customer Service Issues</em>: A telecommunications company applies the same approach to understand recurring customer complaints. Solutions suggested by AI contribute to a 30% reduction in complaints in the first quarter.</p><p>These examples showcase how blending traditional techniques with AI can bring swift and tangible results for modern businesses.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SZZKN0JG63SZGRIToVmQVQ.png" /><figcaption>5 why method</figcaption></figure><p>follow one example bellow :</p><p>Here’s an example of the 5 Whys process and the solutions that came up from the analysis.</p><ol><li><em>Why was the hospital’s new health care app delayed by 4 weeks?</em><br> <strong>Answer</strong>: Additional complexities were revealed in the testing phase.</li><li><em>Why did additional difficulties come up?</em><br> <strong>Answer</strong>: The creation phase did not include a solution for a major use case.</li><li><em>Why did the creation phase miss a major use case?</em><br> <strong>Answer</strong>: The original brainstorming sessions did not include employees from the department.</li><li><em>Why were key employees missing from the brainstorming sessions?</em><br> <strong>Answer</strong>: The brainstorming session was led by a particular department. In this case, additional departments were brought in later in the development process.</li><li><em>Why is this process managed by a particular department?<br> </em><strong>Answer</strong>: This is how it’s always been done.</li></ol><p>Solutions found in the team:</p><blockquote><em>Create a new process that builds on the strengths of all departments and encourages interdepartmental collaboration. Ensure key employees from every department are present in all brainstorming sessions throughout all phases.</em></blockquote><h4><strong>Let’s chek the ai model sugestion :</strong></h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/834/1*GdRJ38Q1nhWQ2sSlt1WOTg.png" /><figcaption>ai model sugestion based 5 why input</figcaption></figure><p>When compared to the model’s suggestions, a clear difference in analytical depth emerges. The model not only aligned with the team’s emphasis on collaboration but went steps further. It presented a comprehensive analysis of potential causes and offered a myriad of solutions. From advocating for rigorous testing and regular updates to considering agile methodologies and equipping the development team with resources, the model showcased an expansive approach.</p><p>While both the team and the model value interdepartmental collaboration, it’s evident that the model provides a more detailed and extensive set of potential solutions. This comparison underlines the model’s exceptional capability in problem analysis and solution generation.</p><p>The team’s insights provided a solid foundation for addressing organizational challenges. When combined with the model’s detailed analysis, we achieve a more comprehensive solution. Together, human expertise and the model’s analytical capabilities create a powerful synergy for enhanced decision-making.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z3BGysJXQZpe-FJpDHkHzg.gif" /></figure><p>Don’t lag in this ever-evolving world. It’s time to embrace the revolution and see how AI can elevate your problem analysis to new heights. Want to witness the magic? Try out our tool now and feel the transformation in your operations.</p><p>👉<a href="https://action-plan-git-main-bernardcaldas.vercel.app/">https://action-plan-git-main-bernardcaldas.vercel.app/</a></p><p>See our tool in action below!”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1002/1*hKPrqHnOSE6Abgdt-6bWdA.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6bd6811b87ce" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/elevating-rca-to-new-heights-the-fusion-of-ai-and-root-cause-analysis-6bd6811b87ce">Elevating RCA and 5 why method to New Heights: The Fusion of AI and Root Cause Analysis</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Revolucionando o 5W2H e 5 Porquês com Inteligência Artificial]]></title>
            <link>https://bernardo-acaldas.medium.com/revolucionando-o-5w2h-e-5-porqu%C3%AAs-com-intelig%C3%AAncia-artificial-adbf47c489fd?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/adbf47c489fd</guid>
            <category><![CDATA[causa-raiz]]></category>
            <category><![CDATA[5w2h]]></category>
            <category><![CDATA[engenharia-de-produção]]></category>
            <category><![CDATA[5ways]]></category>
            <category><![CDATA[chatgpt]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Tue, 15 Aug 2023 14:25:25 GMT</pubDate>
            <atom:updated>2025-01-29T14:42:48.108Z</atom:updated>
            <content:encoded><![CDATA[<p>Como a IA Está Revolucionando a Análise de Problemas nas Empresas</p><p>link gratuito para leitura do artigo completo: <a href="https://bernardo-acaldas.medium.com/revolucionando-o-5w2h-e-5-porqu%C3%AAs-com-intelig%C3%AAncia-artificial-adbf47c489fd?source=friends_link&amp;sk=b8617d79b6ea23257000112b2e5a8f61">https://bernardo-acaldas.medium.com/revolucionando-o-5w2h-e-5-porqu%C3%AAs-com-intelig%C3%AAncia-artificial-adbf47c489fd?source=friends_link&amp;sk=b8617d79b6ea23257000112b2e5a8f61</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*SsHZiy87Cu0fqWjJISIBUA.png" /></figure><p>No cenário empresarial atual, identificar e resolver problemas de forma eficiente é mais crucial do que nunca. Ferramentas tradicionais como o <strong>5W2H</strong> e os <strong>5 Porquês</strong> têm sido pilares nesse processo, fornecendo estruturas sólidas para compreensão e investigação profunda. Mas em uma era dominada pela transformação digital, como podemos elevar essas metodologias a um novo patamar?</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*wbnA70SNXH2eZc57XRserQ.png" /></figure><p><strong>As Limitações dos Métodos Tradicionais</strong></p><p>Embora eficazes, essas ferramentas podem ser limitadas pela subjetividade humana e pela quantidade de tempo necessária para uma análise completa. É aqui que a Inteligência Artificial entra em jogo.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/714/1*wd_WkUiEG86ENkkqcDVu9A.png" /><figcaption>exemplo de formulário usado hoje em dia</figcaption></figure><p><strong>A Integração da IA com o 5W2H e os 5 Porquês</strong></p><p>Ao incorporar IA, o tradicional formulário 5W2H e 5 Porquês torna-se uma ferramenta dinâmica e inteligente. A solução apresentada não apenas sugere respostas baseadas em dados reais, mas também otimiza essas respostas, proporcionando insights mais profundos e ações mais eficazes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*ARKCZvvu-oASQBlnSR7jWA.png" /><figcaption>Aplicação em funcionamento sugerindo uma ação com base nas respostas do formulário.</figcaption></figure><p><strong>Casos de Uso em Empresas: Resultados Reais para Empresas Modernas</strong></p><ol><li><strong>Melhoria de Processos:</strong> Uma fábrica de automóveis utiliza o formulário 5W2H e 5 Porquês potencializado pela IA para identificar gargalos na linha de produção. Com sugestões inteligentes, a empresa consegue aumentar a eficiência em 20% em apenas algumas semanas.</li><li><strong>Resolução de Problemas de Atendimento ao Cliente:</strong> Uma empresa de telecomunicações aplica a mesma abordagem para entender as reclamações recorrentes dos clientes. As soluções sugeridas pela IA ajudam na redução das queixas em 30% no primeiro trimestre.</li></ol><p>Esses exemplos ilustram como a fusão de técnicas tradicionais com a IA pode trazer resultados tangíveis e rápidos para as empresas modernas.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*z3BGysJXQZpe-FJpDHkHzg.gif" /><figcaption>Gif website</figcaption></figure><p><strong>Exemplo Prático: Nosso Modelo em Ação</strong></p><p>Para ilustrar esse potencial, criamos um modelo funcional que você pode testar. Embora não seja um produto final, ele serve como uma amostra do que é possível alcançar ao combinar técnicas tradicionais com inteligência artificial.</p><p>👉 <a href="https://action-plan-git-main-bernardcaldas.vercel.app/">Experimente nosso modelo aqui</a> e veja por si mesmo como a IA pode transformar a análise de problemas.</p><p><strong>Casos de Uso Potenciais em Empresas Modernas</strong></p><ul><li><strong>Melhoria de Processos na Indústria</strong></li><li>Empresas podem utilizar essa abordagem para identificar gargalos e otimizar operações, potencialmente aumentando a eficiência e reduzindo custos.</li><li><strong>Aprimoramento do Atendimento ao Cliente</strong></li><li>Organizações de serviços podem aprofundar a compreensão das reclamações dos clientes e implementar soluções mais eficazes, melhorando a satisfação e a fidelidade.</li></ul><p><strong>Participe Desta Revolução</strong></p><p>Convidamos você a explorar nosso modelo e compartilhar seus feedbacks. Sua contribuição é valiosa para aprimorarmos e desenvolvermos soluções que realmente façam a diferença no mundo empresarial.</p><p>👉<a href="https://action-plan-git-main-bernardcaldas.vercel.app/">https://action-plan-git-main-bernardcaldas.vercel.app/</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=adbf47c489fd" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Analisar, extrair e armazenar dados a partir de imagens]]></title>
            <link>https://medium.com/data-hackers/analisar-extrair-e-armazenar-dados-a-partir-de-imagens-6c4896472b06?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/6c4896472b06</guid>
            <category><![CDATA[inteligencia-artificial]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[face-detection-app]]></category>
            <category><![CDATA[visão-computacional]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Sat, 08 May 2021 15:03:31 GMT</pubDate>
            <atom:updated>2021-05-18T17:01:28.605Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*TT3OufZruiDre4uz" /><figcaption>Photo by <a href="https://unsplash.com/@spacex?utm_source=medium&amp;utm_medium=referral">SpaceX</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>Neste post irei mostrar como extrair dados e extrair informações importantes para futuras análises a partir de fotos de rostos inicialmente.</p><p>para realizar as análises foi utilizado o framework em python chamado DeepFace: <a href="https://github.com/serengil/deepface">https://github.com/serengil/deepface</a></p><p>Em outro post foi mostrado como identificar fotos em em imagens, segue link :</p><p><a href="https://medium.com/data-hackers/webapp-para-detectar-faces-8ee83a1442be">https://medium.com/data-hackers/webapp-para-detectar-faces-8ee83a1442be</a></p><p>Porém seria interessante além de armazenar e contar os rostos extrair mais dados dessas imagens como;</p><ul><li>Gênero</li><li>Idade</li><li>Raça</li><li>Humor</li></ul><p>Para explicar melhor o que pretendemos fazer segue abaixo vídeo;</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FzpczJ6nuO_4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DzpczJ6nuO_4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FzpczJ6nuO_4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/00dc1e76a6f6e765483581b7c1b6624a/href">https://medium.com/media/00dc1e76a6f6e765483581b7c1b6624a/href</a></iframe><h3>Face detector</h3><p>Para fazer a detecção dos rostos em imagens usamos MTCNN ( <a href="https://arxiv.org/abs/1604.02878"><em>Multi-Task Cascaded Convolutional Neural Network</em></a><em>) </em>igual no projeto anterior, <em>porque MTCNN</em> ? porque os resultados são melhores que outras tecnologias para detecção que foram testadas como ( haar cascades, dlib )</p><p>Para realizar a extração dos rostos em uma imagem foi usado a função abaixo;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/472/1*dW4SGDRzoaoHlZUEHs2f_Q.png" /></figure><p>após cada detecção iremos salvar cada detecção foto em um diretório local para realizar a análises dos atributos descritos;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/374/1*Vqu3VQ8TYXd2b0VHUPwi_g.png" /><figcaption>salvando detecções individualmente</figcaption></figure><h3>Reconhecimento Facial</h3><p>Para realizar a análise dos atributos para cada foto de rosto salva usamos um framework chamado <a href="https://github.com/serengil/deepface">deepFace</a></p><pre>img = &#39;sample.png&#39;</pre><pre>attributes = [&#39;age&#39;, &#39;gender&#39;, &#39;race&#39;, &#39;emotion&#39;]</pre><pre>demography = DeepFace.analyze(img, attributes)</pre><pre>st.write(demography[&#39;age&#39;])</pre><pre>st.write(demography[&#39;dominant_emotion&#39;])</pre><pre>st.write(demography[&#39;gender&#39;])</pre><pre>st.write(demography[&#39;dominant_race&#39;])</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/734/1*OPcPUm3f8CsgtvR_P4u-tA.png" /><figcaption>exemplo saída deepFace</figcaption></figure><p>Agora que temos as análises para cada foto iremos salvar essas informações em um dicionário, usamos um serviço gratuito chamado firebase, onde é possível usar funcionalidades do ‘storage’ para armazenar arquivos de fotos e o serviço ‘realtime-database’, para armazenar dados nosql.</p><h3>Armazenamento</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*-ajpvDVk1sHl089wOEazng.png" /><figcaption>firebase</figcaption></figure><blockquote>O serviço em nuvem para desenvolvedores móveis é um back-end completo para aplicações mobile (Android e iOS) e aplicações web. Com visual limpo e de uso simples, o Firebase é uma plataforma dedicada e SDK para a construção de aplicativos. Atualmente, o serviço suporta desenvolvimento nas linguagens de programação C++, Java, Javascript, Node.js, Objective-C e Swift.</blockquote><p>incluindo informações no banco ex ;</p><pre>data = {&quot;name&quot;: &quot;Mortimer &#39;Morty&#39; Smith&quot;}<br>db.child(&quot;users&quot;).push(data)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*1dMlKLpVhK1_mXLAzX4Z6A.png" /><figcaption>realtime database</figcaption></figure><p>incluindo imagens e arquivos no banco</p><pre>for i in imageList:</pre><pre>storage.child(str(i)).put(i)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/576/1*LG_Gz9lOmoxbdNqDBe6HdQ.png" /><figcaption>storage</figcaption></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FsuymzXyDbuEPC%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fspongebob-squarepants-surprised-patrick-psychedelic-suymzXyDbuEPC&amp;image=https%3A%2F%2Fmedia2.giphy.com%2Fmedia%2FsuymzXyDbuEPC%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="261" frameborder="0" scrolling="no"><a href="https://medium.com/media/4c24f867ffbe33c74b4b24ede91dba19/href">https://medium.com/media/4c24f867ffbe33c74b4b24ede91dba19/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=6c4896472b06" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-hackers/analisar-extrair-e-armazenar-dados-a-partir-de-imagens-6c4896472b06">Analisar, extrair e armazenar dados a partir de imagens</a> was originally published in <a href="https://medium.com/data-hackers">Data Hackers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Extract and store data from images]]></title>
            <link>https://medium.com/analytics-vidhya/extract-and-store-data-from-images-c2737c860428?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c2737c860428</guid>
            <category><![CDATA[firebase]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[cloud-storage]]></category>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[face-recognition]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Sun, 11 Apr 2021 21:54:54 GMT</pubDate>
            <atom:updated>2021-04-12T15:48:56.158Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*K5eYjSq1DVQsWQGIyUvcoA.png" /></figure><p>In this post we are using firebase to store our data collected for the application such as gender, age and dominant emotion from each detection.</p><p>before discuss please follow video bellow with results of the application we deploy using streamlit | python and firebase.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FzpczJ6nuO_4%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DzpczJ6nuO_4&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FzpczJ6nuO_4%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/00dc1e76a6f6e765483581b7c1b6624a/href">https://medium.com/media/00dc1e76a6f6e765483581b7c1b6624a/href</a></iframe><h3>Emotion recognition</h3><p>In the <a href="https://en.wikipedia.org/wiki/18th_century">18th</a> and <a href="https://en.wikipedia.org/wiki/19th_century">19th century</a> the belief that facial expressions revealed the moral worth or true inner state of a human was widespread and <a href="https://en.wikipedia.org/wiki/Physiognomy">physiognomy</a> was a respected <a href="https://en.wikipedia.org/wiki/Science">science</a> in the <a href="https://en.wikipedia.org/wiki/Western_world">Western world</a>. From the early 19th century onwards <a href="https://en.wikipedia.org/wiki/Photography">photography</a> was used in the physiognomic analysis of facial features and facial expression to detect <a href="https://en.wikipedia.org/wiki/Insanity">insanity</a> and <a href="https://en.wikipedia.org/wiki/Dementia">dementia</a>.<a href="https://en.wikipedia.org/wiki/Facial_recognition_system#cite_note-159">[159]</a> In the 1960s and 1970s the study of human emotions and its expressions was reinvented by <a href="https://en.wikipedia.org/wiki/Psychologist">psychologists</a>, who tried to define a normal range of emotional responses to events.<a href="https://en.wikipedia.org/wiki/Facial_recognition_system#cite_note-160">[160]</a> The research on automated <a href="https://en.wikipedia.org/wiki/Emotion_recognition">emotion recognition</a> has since the 1970s focused on <a href="https://en.wikipedia.org/wiki/Facial_expression">facial expressions</a> and <a href="https://en.wikipedia.org/wiki/Speech">speech</a>, which are regarded as the two most important ways in which humans communicate <a href="https://en.wikipedia.org/wiki/Emotions">emotions</a> to other humans. In the 1970s the <a href="https://en.wikipedia.org/wiki/Facial_Action_Coding_System">Facial Action Coding System</a> (FACS) categorization for the physical expression of emotions was established.<a href="https://en.wikipedia.org/wiki/Facial_recognition_system#cite_note-161">[161]</a> Its developer <a href="https://en.wikipedia.org/wiki/Paul_Ekman">Paul Ekman</a> maintains that there are six emotions that are universal to all human beings and that these can be coded in facial expressions.<a href="https://en.wikipedia.org/wiki/Facial_recognition_system#cite_note-162">[162]</a> Research into automatic emotion specific expression recognition has in the past decades focused on frontal view images of human faces.<a href="https://en.wikipedia.org/wiki/Facial_recognition_system#cite_note-163">[163]</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*STB2G7oWd3zv_wo-3-gpqg.jpeg" /></figure><h4>Face Detectors</h4><blockquote>Face detection and alignment are early stages of a modern face recognition pipeline. <a href="https://sefiks.com/2020/02/23/face-alignment-for-face-recognition-in-python-within-opencv/">OpenCV</a>, <a href="https://sefiks.com/2020/08/25/deep-face-detection-with-opencv-in-python/">SSD</a>, <a href="https://sefiks.com/2020/07/11/face-recognition-with-dlib-in-python/">Dlib</a> and <a href="https://sefiks.com/2020/09/09/deep-face-detection-with-mtcnn-in-python/">MTCNN</a> methods are wrapped in deepface as a detector. You can optionally pass a custom detector to functions in deepface interface. MTCNN is the default detector if you won&#39;t pass any detector.</blockquote><p>source : <a href="https://github.com/serengil/deepface">https://github.com/serengil/deepface</a></p><pre>backends = [&#39;opencv&#39;, &#39;ssd&#39;, &#39;dlib&#39;, &#39;mtcnn&#39;]</pre><pre>for backend in backends:<br>   <br>  demography = DeepFace.analyze(&quot;img4.jpg&quot;, detector_backend =             backend)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*Ep_0-isQEHPurubUXoITVw.jpeg" /></figure><h3>Storage</h3><blockquote>The Firebase Realtime Database is a cloud-hosted database. Data is stored as JSON and synchronized in realtime to every connected client. When you build cross-platform apps with our iOS, Android, and JavaScript SDKs, all of your clients share one Realtime Database instance and automatically receive updates with the newest data.</blockquote><p>source: <a href="https://firebase.google.com/docs/database">https://firebase.google.com/docs/database</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/736/1*h9PwZ0EovHTZ4RQeCwpvZw.png" /></figure><h3>Storage</h3><p>The storage service allows you to upload images to Firebase.</p><h3>child</h3><p>Just like with the Database service, you can build paths to your data with the Storage service.</p><pre>storage.child(&quot;images/example.jpg&quot;)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ofghQU4n55eva1TfELWQIw.png" /><figcaption>Storage Firebase service ( files captured)</figcaption></figure><h3>Save Data</h3><h4>push</h4><p>To save data with a unique, auto-generated, timestamp-based key, use the push() method.</p><pre>data = {&quot;name&quot;: &quot;Mortimer &#39;Morty&#39; Smith&quot;}<br>db.child(&quot;users&quot;).push(data)</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*DiaZpEvKZaRP71rWQLvSCw.png" /><figcaption>Realtime Database Firebase service ( data from images )</figcaption></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FTQAdxJWqM5soE%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FTQAdxJWqM5soE%2Fgiphy.gif&amp;image=https%3A%2F%2Fi.giphy.com%2Fmedia%2FTQAdxJWqM5soE%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="326" frameborder="0" scrolling="no"><a href="https://medium.com/media/7526d59b8bb8ade3afe09befa0a48efa/href">https://medium.com/media/7526d59b8bb8ade3afe09befa0a48efa/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c2737c860428" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/extract-and-store-data-from-images-c2737c860428">Extract and store data from images</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How create Image Recomendation system]]></title>
            <link>https://medium.com/analytics-vidhya/how-create-image-recomendation-system-3dcc5edf1597?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/3dcc5edf1597</guid>
            <category><![CDATA[recommender-systems]]></category>
            <category><![CDATA[cosine-similarity]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[recommendation-system]]></category>
            <category><![CDATA[keras]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Sun, 06 Dec 2020 19:36:17 GMT</pubDate>
            <atom:updated>2020-12-07T15:07:47.860Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/757/1*dCEBhlDpuHkwEErkprmG_g.jpeg" /></figure><p>A <strong>recommender system</strong>, or a <strong>recommendation system</strong> (sometimes replacing ‘system’ with a synonym such as platform or engine), is a subclass of <a href="https://en.wikipedia.org/wiki/Information_filtering_system">information filtering system</a> that seeks to predict the “rating” or “preference” a user would give to an item.<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-handbook-1">[1]</a><a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-2">[2]</a> They are primarily used in commercial applications.</p><p>Recommender systems are utilized in a variety of areas and are most commonly recognized as playlist generators for video and music services, product recommenders for online stores, or content recommenders for social media platforms and open web content recommenders.<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-twitterwtf2-3">[3]</a><a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-4">[4]</a> These systems can operate using a single input, like music, or multiple inputs within and across platforms like news, books, and search queries. There are also popular recommender systems for specific topics like restaurants and <a href="https://en.wikipedia.org/wiki/Online_dating">online dating</a>. Recommender systems have also been developed to explore research articles and experts,<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-expertseer-5">[5]</a> collaborators,<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-collabseer-6">[6]</a> and financial services.</p><p>( <a href="https://en.wikipedia.org/wiki/Recommender_system">https://en.wikipedia.org/wiki/Recommender_system</a> )</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*E8c4PEwsogQQWJPErGda2A.gif" /></figure><h3>Collaborative filtering</h3><p>A key advantage of the collaborative filtering approach is that it does not rely on machine analyzable content and therefore it is capable of accurately recommending complex items such as movies without requiring an “understanding” of the item itself. Many algorithms have been used in measuring user similarity or item similarity in recommender systems. For example, the <a href="https://en.wikipedia.org/wiki/K-nearest_neighbors_algorithm">k-nearest neighbor</a> (k-NN) approach<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-38">[38]</a> and the <a href="https://en.wikipedia.org/wiki/Pearson_correlation">Pearson Correlation</a> as first implemented by Allen.<a href="https://en.wikipedia.org/wiki/Recommender_system#cite_note-39">[39]</a></p><p>When building a model from a user’s behavior, a distinction is often made between explicit and <a href="https://en.wikipedia.org/wiki/Implicit_data_collection">implicit</a> forms of <a href="https://en.wikipedia.org/wiki/Data_collection">data collection</a>.</p><p>Examples of explicit data collection include the following:</p><ul><li>Asking a user to rate an item on a sliding scale.</li><li>Asking a user to search.</li><li>Asking a user to rank a collection of items from favorite to least favorite.</li><li>Presenting two items to a user and asking him/her to choose the better one of them.</li><li>Asking a user to create a list of items that he/she likes (see <a href="https://en.wikipedia.org/wiki/Rocchio_algorithm"><em>Rocchio classification</em></a> or other similar techniques).</li></ul><h3>Cosine similarity</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/697/1*_Bf9goaALQrS_0XkBozEiQ.png" /></figure><p>Using Keras and CNN vgg16 we going to develop a algorithm to recommend similar products ;</p><pre><em># imports</em><br><br>from keras.applications import vgg16<br>from keras.preprocessing.image import load_img,img_to_array<br>from keras.models import Model<br>from keras.applications.imagenet_utils import preprocess_input<br><br>from PIL import Image<br>import os<br>import matplotlib.pyplot as plt<br>import numpy as np<br>from sklearn.metrics.pairwise import cosine_similarity<br>import pandas as pd</pre><pre><em># parameters setup</em><br><br>imgs_path = &quot;../input/style/&quot;<br>imgs_model_width, imgs_model_height = 224, 224<br><br>nb_closest_images = 5 <em># number of most similar images to retrieve</em></pre><h3>1. load the VGG pre-trained model from Keras</h3><p>Keras module contains several pre-trained models that can be loaded very easily.</p><p>For our recommender system based on visual similarity, we need to load a Convolutional Neural Network (CNN) that will be able to interpret the image contents.</p><p>In this example we will load the VGG16 model trained on imagenet, a big labeled images database.</p><p>If we take the whole model, we will get an output containing probabilities to belong to certain classes, but that is not what we want.</p><p>We want to retrieve all the information that the model was able to get in the images.</p><p>In order to do so, we have to remove the last layers of the CNN which are only used for classes predictions.</p><pre>files = [imgs_path + x for x in os.listdir(imgs_path) if “jpg” in x]</pre><pre>print(“number of images:”,len(files))</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/588/1*53q-U_dhMZwNoj5CP0VsTA.png" /><figcaption>example path files using in project</figcaption></figure><pre><em># load the model</em><br>vgg_model = vgg16.VGG16(weights=&#39;imagenet&#39;)<br><br><em># remove the last layers in order to get features instead of predictions</em><br>feat_extractor = Model(inputs=vgg_model.input, outputs=vgg_model.get_layer(&quot;fc2&quot;).output)<br><br><em># print the layers of the CNN</em><br>feat_extractor.summary()</pre><pre># compute cosine similarities between images</pre><pre>cosSimilarities = cosine_similarity(imgs_features)</pre><pre># store the results into a pandas dataframe</pre><pre>cos_similarities_df = pd.DataFrame(cosSimilarities, columns=files, index=files)<br>cos_similarities_df.head()</pre><pre>def retrieve_most_similar_products(given_img):</pre><pre>print(&quot;--------------------------------------------------------&quot;)<br>    print(&quot;original product:&quot;)</pre><pre>original = load_img(given_img, target_size=(imgs_model_width, imgs_model_height))<br>    plt.imshow(original)<br>    plt.show()</pre><pre>print(&quot;-------------------------------------------------------&quot;)<br>    print(&quot;most similar products:&quot;)</pre><pre>closest_imgs = cos_similarities_df[given_img].sort_values(ascending=False)[1:nb_closest_images+1].index<br>    closest_imgs_scores = cos_similarities_df[given_img].sort_values(ascending=False)[1:nb_closest_images+1]</pre><pre>for i in range(0,len(closest_imgs)):<br>        original = load_img(closest_imgs[i], target_size=(imgs_model_width, imgs_model_height))<br>        plt.imshow(original)<br>        plt.show()<br>        print(&quot;similarity score : &quot;,closest_imgs_scores[i])</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/723/1*Q962VmSpuXz5MIFAOsIDQQ.png" /><figcaption>results</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/745/1*gKtG65yu8kRmP6PXfgKT7g.png" /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FdQNP6OnyFUePu%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fmedia.giphy.com%2Fmedia%2FdQNP6OnyFUePu%2Fgiphy.gif&amp;image=https%3A%2F%2Fi.giphy.com%2Fmedia%2FdQNP6OnyFUePu%2Fgiphy-downsized-large.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="326" frameborder="0" scrolling="no"><a href="https://medium.com/media/5cda4cddf4f16947e9c45aa231946055/href">https://medium.com/media/5cda4cddf4f16947e9c45aa231946055/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3dcc5edf1597" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/how-create-image-recomendation-system-3dcc5edf1597">How create Image Recomendation system</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[WebApp para detectar faces]]></title>
            <link>https://medium.com/data-hackers/webapp-para-detectar-faces-8ee83a1442be?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/8ee83a1442be</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[face]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[visão-computacional]]></category>
            <category><![CDATA[rosto]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Sun, 08 Nov 2020 16:17:25 GMT</pubDate>
            <atom:updated>2021-04-25T14:54:51.084Z</atom:updated>
            <content:encoded><![CDATA[<p>Neste post irei mostrar como fazer um webApp para detectar rostos em fotos usando deep learning e streamlit.io</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*FD1zwMNJKFhG14CwOd0u4w.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/310/1*72jJj9Qh42brAgEaALHMMA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*iEq6DjCw4QMmBkooDfMfZg.jpeg" /><figcaption>MTCNN sample</figcaption></figure><p><strong>Detecção facial</strong> é uma técnica de <a href="https://pt.wikipedia.org/wiki/Processamento_de_imagem">processamento de imagem</a> e <a href="https://pt.wikipedia.org/wiki/Vis%C3%A3o_computacional">visão computacional</a> para determinar a existência, ou não, de faces numa determinada imagem e, caso exista(m), retornar a localização da(s) mesma(s). <a href="https://pt.wikipedia.org/wiki/Detec%C3%A7%C3%A3o_facial#cite_note-:0-1">[1]</a> Apesar de ser uma tarefa trivial para seres-humanos, é um problema desafiador para computadores, visto que rostos podem variar em cor, iluminação, posicionamento e escala, dificultando a detecção automática. <a href="https://pt.wikipedia.org/wiki/Detec%C3%A7%C3%A3o_facial#cite_note-:0-1">[1]</a></p><p>A solução para o problema de detecção facial engloba diversas técnicas, como segmentação, extração, e verificação de faces e de suas características em um plano de fundo não controlado. <a href="https://pt.wikipedia.org/wiki/Detec%C3%A7%C3%A3o_facial#cite_note-:1-2">[2]</a> O processo de detecção facial pode servir como base para muitos sistemas, como por exemplo, sistemas de reconhecimento facial. — <a href="https://pt.wikipedia.org/wiki/Detec%C3%A7%C3%A3o_facial">https://pt.wikipedia.org/wiki/Detec%C3%A7%C3%A3o_facial</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*QrQhcdGIEwEH6kRmZAecgg.gif" /><figcaption>sample gif, reconhecimento facial</figcaption></figure><p>O reconhecimento de rosto pode ser dividido em várias etapas. A imagem abaixo mostra um exemplo de pipeline de reconhecimento facial.</p><p>1. Detecção de rosto — Detectar um ou mais rostos em uma imagem.<br>2. Extração de recursos — Extração dos recursos mais importantes de uma imagem do rosto.<br>3. Classificação de rosto — Classifica o rosto com base nos recursos extraídos.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k1YBPFSjsZPAJFQGxsFO9w.png" /></figure><h3>MTCNN</h3><p>Implementação do detector facial MTCNN para Keras em Python3.4 +. Ele foi escrito do zero, usando como referência a implementação do MTCNN de David Sandberg (<a href="https://github.com/davidsandberg/facenet/tree/master/src/align">FaceNet’s MTCNN</a>) usando Facenet. Baseado no paper <em>Zhang, K et al. (2016)</em> <a href="https://pypi.org/project/mtcnn/#zhang2016">[ZHANG2016]</a>.</p><p>fonte = <a href="https://pypi.org/project/mtcnn/">https://pypi.org/project/mtcnn/</a></p><pre>&gt;&gt;&gt; from mtcnn import MTCNN<br>&gt;&gt;&gt; import cv2<br>&gt;&gt;&gt;<br>&gt;&gt;&gt; img = cv2.cvtColor(cv2.imread(&quot;ivan.jpg&quot;), cv2.COLOR_BGR2RGB)<br>&gt;&gt;&gt; detector = MTCNN()<br>&gt;&gt;&gt; detector.detect_faces(img)<br>[<br>    {<br>        &#39;box&#39;: [277, 90, 48, 63],<br>        &#39;keypoints&#39;:<br>        {<br>            &#39;nose&#39;: (303, 131),<br>            &#39;mouth_right&#39;: (313, 141),<br>            &#39;right_eye&#39;: (314, 114),<br>            &#39;left_eye&#39;: (291, 117),<br>            &#39;mouth_left&#39;: (296, 143)<br>        },<br>        &#39;confidence&#39;: 0.99851983785629272<br>    }<br>]</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*iEq6DjCw4QMmBkooDfMfZg.jpeg" /><figcaption>exemplo output usando MTCNN</figcaption></figure><p>Usando Heroku realizamos o deploy do projeto para testes;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/310/1*72jJj9Qh42brAgEaALHMMA.png" /></figure><h4>Link para acessar o webbApp :</h4><p>Link : <a href="https://webapp-facedetect.herokuapp.com/">https://webapp-facedetect.herokuapp.com/</a></p><h4>Youtube video tutorial</h4><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F9szgVzUtCcA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D9szgVzUtCcA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F9szgVzUtCcA%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/12bdc8529f3d6483f3da3472f42fa295/href">https://medium.com/media/12bdc8529f3d6483f3da3472f42fa295/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/769/1*O7d89EoKCjgr0jG4otcutw.png" /><figcaption>captura de tela teste com aplicação</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/725/1*JZzyIcorLNLm4HhITBPWAw.png" /><figcaption>amostra resultados da aplicação</figcaption></figure><p><a href="https://webapp-facedetect.herokuapp.com/">Streamlit</a></p><p>Link : <a href="https://webapp-facedetect.herokuapp.com/">https://webapp-facedetect.herokuapp.com/</a></p><p>código fonte do projeto : <a href="https://github.com/bernardcaldas/webapp-face">https://github.com/bernardcaldas/webapp-face</a></p><p><a href="https://github.com/bernardcaldas/webapp-face">bernardcaldas/webapp-face</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*TGOufhH-yUZIuzmlEjo-LQ.gif" /></figure><p>twitter : <a href="https://twitter.com/bernardocalda10">https://twitter.com/bernardocalda10</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8ee83a1442be" width="1" height="1" alt=""><hr><p><a href="https://medium.com/data-hackers/webapp-para-detectar-faces-8ee83a1442be">WebApp para detectar faces</a> was originally published in <a href="https://medium.com/data-hackers">Data Hackers</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Face detection webApp]]></title>
            <link>https://medium.com/analytics-vidhya/face-detection-webapp-5f947cffbfcb?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/5f947cffbfcb</guid>
            <category><![CDATA[deep-learning]]></category>
            <category><![CDATA[facedetection]]></category>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[mtcnn]]></category>
            <category><![CDATA[face-net]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Sun, 25 Oct 2020 18:11:05 GMT</pubDate>
            <atom:updated>2021-04-26T18:12:17.955Z</atom:updated>
            <content:encoded><![CDATA[<h3>Face Detection webApp</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/620/1*FD1zwMNJKFhG14CwOd0u4w.jpeg" /></figure><p>In this quick post I wanted to share a very popular way to deploy your model as webApp using <a href="https://www.streamlit.io/">Streamlit</a>.io</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*8_tOPp_41jV_3X8RE4E7fw.png" /></figure><p><strong>“ Face detection</strong> is a computer technology being used in a variety of applications that identifies human faces in digital images.<a href="https://en.wikipedia.org/wiki/Face_detection#cite_note-1">[1]</a> Face detection also refers to the psychological process by which humans locate and attend to faces in a visual scene.” — <a href="https://en.wikipedia.org/wiki/Face_detection">https://en.wikipedia.org/wiki/Face_detection</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*k1YBPFSjsZPAJFQGxsFO9w.png" /></figure><p>Formally, Face Recognition is defined as <em>the problem of identifying or verifying faces in an image</em>. How exactly do we recognize a face in an image?</p><p>Face recognition can be divided into multiple steps. The image below shows an example of a face recognition pipeline.</p><ol><li>Face detection — Detecting one or more faces in an image.</li><li>Feature extraction — Extracting the most important features from an image of the face.</li><li>Face classification — Classifying the face based on extracted features.</li></ol><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*QrQhcdGIEwEH6kRmZAecgg.gif" /><figcaption>sample face detection</figcaption></figure><p>There are various ways to implement each of the steps in a face recognition pipeline. In this post we’ll focus on popular deep learning approaches where we perform face detection using MTCNN, feature extraction using FaceNet and classification using Softmax. — <a href="https://arsfutura.com/magazine/face-recognition-with-facenet-and-mtcnn/">https://arsfutura.com/magazine/face-recognition-with-facenet-and-mtcnn/</a></p><h3>MTCNN</h3><p>Implementation of the MTCNN face detector for Keras in Python3.4+. It is written from scratch, using as a reference the implementation of MTCNN from David Sandberg (<a href="https://github.com/davidsandberg/facenet/tree/master/src/align">FaceNet’s MTCNN</a>) in Facenet. It is based on the paper <em>Zhang, K et al. (2016)</em> <a href="https://pypi.org/project/mtcnn/#zhang2016">[ZHANG2016]</a>.</p><p>source = <a href="https://pypi.org/project/mtcnn/">https://pypi.org/project/mtcnn/</a></p><pre>&gt;&gt;&gt; from mtcnn import MTCNN<br>&gt;&gt;&gt; import cv2<br>&gt;&gt;&gt;<br>&gt;&gt;&gt; img = cv2.cvtColor(cv2.imread(&quot;ivan.jpg&quot;), cv2.COLOR_BGR2RGB)<br>&gt;&gt;&gt; detector = MTCNN()<br>&gt;&gt;&gt; detector.detect_faces(img)<br>[<br>    {<br>        &#39;box&#39;: [277, 90, 48, 63],<br>        &#39;keypoints&#39;:<br>        {<br>            &#39;nose&#39;: (303, 131),<br>            &#39;mouth_right&#39;: (313, 141),<br>            &#39;right_eye&#39;: (314, 114),<br>            &#39;left_eye&#39;: (291, 117),<br>            &#39;mouth_left&#39;: (296, 143)<br>        },<br>        &#39;confidence&#39;: 0.99851983785629272<br>    }<br>]</pre><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CzoAg4Shvv6crZuA4Opa5Q.jpeg" /><figcaption>MTCNN output sample</figcaption></figure><h3>Deploy the webApp</h3><p>Deploying a model simply means into an existing production environment which can take in an input and return an output that can be used in making practical business decisions. Here is where Streamlit comes to play !</p><p>follow the webapp, feel free to test ;-)</p><iframe src="" width="0" height="0" frameborder="0" scrolling="no"><a href="https://medium.com/media/0c2a293d54cf765c3d83685c42002b1c/href">https://medium.com/media/0c2a293d54cf765c3d83685c42002b1c/href</a></iframe><h4>follow the webapp, feel free to test ;-)</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*Cdveta6ovG3b-BEWF4gfDA.png" /></figure><p><a href="https://webapp-facedetect.herokuapp.com/">Streamlit</a></p><h4>Youtube video tutorial</h4><figure><img alt="" src="https://cdn-images-1.medium.com/max/225/1*0fjk0CsAMJKpq_KgV9nSzw.png" /></figure><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2F9szgVzUtCcA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3D9szgVzUtCcA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2F9szgVzUtCcA%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/12bdc8529f3d6483f3da3472f42fa295/href">https://medium.com/media/12bdc8529f3d6483f3da3472f42fa295/href</a></iframe><figure><img alt="" src="https://cdn-images-1.medium.com/max/769/1*O7d89EoKCjgr0jG4otcutw.png" /></figure><p>sample webapp</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/725/1*JZzyIcorLNLm4HhITBPWAw.png" /><figcaption>results webapp</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/480/1*TGOufhH-yUZIuzmlEjo-LQ.gif" /></figure><p>See you there ;-)</p><p><a href="https://twitter.com/bernardocalda10">https://twitter.com/bernardocalda10</a></p><p><a href="https://github.com/bernardcaldas">bernardcaldas - Overview</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=5f947cffbfcb" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/face-detection-webapp-5f947cffbfcb">Face detection webApp</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Custom Object detection using ImageAi with few steps]]></title>
            <link>https://medium.com/analytics-vidhya/custom-object-detection-using-imageai-with-few-steps-e97b999bae79?source=rss-f1a3625e122c------2</link>
            <guid isPermaLink="false">https://medium.com/p/e97b999bae79</guid>
            <category><![CDATA[computer-vision]]></category>
            <category><![CDATA[custom-object]]></category>
            <category><![CDATA[beginner]]></category>
            <category><![CDATA[object-detection]]></category>
            <dc:creator><![CDATA[Bernardo Caldas]]></dc:creator>
            <pubDate>Mon, 15 Jun 2020 18:26:45 GMT</pubDate>
            <atom:updated>2020-06-16T13:51:15.067Z</atom:updated>
            <content:encoded><![CDATA[<p>“<strong>ImageAI</strong> is a python library built to empower developers, reseachers and students to build applications and systems with self-contained Deep Learning and Computer Vision capabilities using simple and few lines of code” Official documentation ( <a href="https://imageai.readthedocs.io/en/latest/">https://imageai.readthedocs.io/en/latest/</a>)</p><p>ImageAi is a very powerful tool to create models for custom dataset, practical and easy to use, follow those steps to create your own classifier;</p><p>1- prepare your dataset</p><p>2- Install imageAi</p><p>3- Download pre trained weights ( yolo, RetinaNet )</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/538/1*49565kutSbP-c6dprirKoQ.png" /><figcaption>sample object detection</figcaption></figure><h3>Computer Vision</h3><p>In recent years we can see a lot applications in our life including, autonomous cars, facial detections app, education, military, finance etc.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/0*prQDwnKsOfq_FE9r.gif" /><figcaption>sample object segmentation</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/768/1*8y-mWhuL6oyIQYAA78yXaw.png" /><figcaption>sample object detection</figcaption></figure><h4>Prepare the dataset</h4><p>It is necessary to collect a good number of images for the training process, it is recommended for the project at least 200 images for each class, below follows the structure;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/703/1*jhgHUhyEpSYk0sgK6TGK9w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/421/1*TRfnrAvu2ZJsbuKmj-eShw.png" /></figure><h4>Annotating images</h4><p>ImageAi supports the Pascal Voc annotation format, personally to annotate images I used LabelIMG, an easy to use tool, after the separation seen above the project structure, between training and validation (70% training images and 30% validation images). open LabelImg and we start the process of annotating images as shown below, for each image in the dataset.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*B7_yMg7NV4nOLLRqGmCCEg.png" /></figure><p>after annotating each image the application will generate an xml file for each image that should be directed to the “annotations” folder,</p><h4>Installing ImageAi and training requirements</h4><p>As already described we use Google Colab to perform the training using enabling the GPU to speed up the process;</p><h3>Requirements</h3><p><strong>Python :</strong> <a href="https://python.org/">https://python.org</a></p><p><strong>Tensorflow :</strong> !pip3 install tensorflow ( Versão inferior a 2.0 )</p><p><strong>OpenCV :</strong> !pip3 install opencv-python</p><p><strong>Keras : </strong>!pip3 install keras</p><p><strong>ImageAI : </strong>!pip3 install imageai — upgrade</p><h3>Download pre model yolov3</h3><p><em>!wget </em><a href="https://github.com/OlafenwaMoses/ImageAI/releases/download/essential-v4/pretrained-yolov3.h5"><em>https://github.com/OlafenwaMoses/ImageAI/releases/download/essential-v4/pretrained-yolov3.h5</em></a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/859/1*3rWfICY4nlKTM_l61AJQKA.png" /></figure><p>After executing the previous steps and enabling the notebook’s GPU on google colab we can train our application, for that we must run the code below, the complete project documentation is very well written and easy to understand,</p><p>link: <a href="https://imageai.readthedocs.io/en/latest/customdetection/index.html">https://imageai.readthedocs.io/en/latest/customdetection/index.html</a></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/701/1*eYJuP6lokwWUs6gcopT1fg.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/697/1*1l4NFckCry8lygn78c8jqA.png" /></figure><p>Now just wait for the training process … ( about 3 hours )</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2FtXL4FHPSnVJ0A%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2Fkim-novak-tXL4FHPSnVJ0A&amp;image=https%3A%2F%2Fmedia3.giphy.com%2Fmedia%2FtXL4FHPSnVJ0A%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="287" frameborder="0" scrolling="no"><a href="https://medium.com/media/a103b3e488a7ac962b8cd40079ac2892/href">https://medium.com/media/a103b3e488a7ac962b8cd40079ac2892/href</a></iframe><p>After the end of the training process, we upload a video to test the application using the code below;</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/717/1*rpu5AdfkGjg1R_SqsviLiw.png" /><figcaption>sample for test video object detection</figcaption></figure><p>Follow the results;</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fwww.youtube.com%2Fembed%2FEvqf6hbfxxA%3Ffeature%3Doembed&amp;display_name=YouTube&amp;url=https%3A%2F%2Fwww.youtube.com%2Fwatch%3Fv%3DEvqf6hbfxxA&amp;image=https%3A%2F%2Fi.ytimg.com%2Fvi%2FEvqf6hbfxxA%2Fhqdefault.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=youtube" width="854" height="480" frameborder="0" scrolling="no"><a href="https://medium.com/media/e5a97b0969ba7e3f2edf2a507b3e5d5a/href">https://medium.com/media/e5a97b0969ba7e3f2edf2a507b3e5d5a/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fgiphy.com%2Fembed%2F4PSEQpvV5wUpnmpP1l%2Ftwitter%2Fiframe&amp;display_name=Giphy&amp;url=https%3A%2F%2Fgiphy.com%2Fgifs%2FBounce-TV-cool-ok-thumbs-up-4PSEQpvV5wUpnmpP1l&amp;image=https%3A%2F%2Fmedia1.giphy.com%2Fmedia%2F4PSEQpvV5wUpnmpP1l%2Fgiphy.gif&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=giphy" width="435" height="435" frameborder="0" scrolling="no"><a href="https://medium.com/media/981b8de1d26e0bed8be27c1719f82b1c/href">https://medium.com/media/981b8de1d26e0bed8be27c1719f82b1c/href</a></iframe><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=e97b999bae79" width="1" height="1" alt=""><hr><p><a href="https://medium.com/analytics-vidhya/custom-object-detection-using-imageai-with-few-steps-e97b999bae79">Custom Object detection using ImageAi with few steps</a> was originally published in <a href="https://medium.com/analytics-vidhya">Analytics Vidhya</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>