<?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[testdotai - Medium]]></title>
        <description><![CDATA[Mobile app development, test automation, artificial intelligence and more… - Medium]]></description>
        <link>https://medium.com/testdotai?source=rss----4417152e2fa4---4</link>
        <image>
            <url>https://cdn-images-1.medium.com/proxy/1*TGH72Nnw24QL3iV9IOm4VA.png</url>
            <title>testdotai - Medium</title>
            <link>https://medium.com/testdotai?source=rss----4417152e2fa4---4</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 05:07:05 GMT</lastBuildDate>
        <atom:link href="https://medium.com/feed/testdotai" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Test.ai Has a New Look]]></title>
            <link>https://medium.com/testdotai/test-ai-has-a-new-look-619b351b4f7b?source=rss----4417152e2fa4---4</link>
            <guid isPermaLink="false">https://medium.com/p/619b351b4f7b</guid>
            <category><![CDATA[startup]]></category>
            <category><![CDATA[design]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[branding]]></category>
            <category><![CDATA[logo-design]]></category>
            <dc:creator><![CDATA[test.ai]]></dc:creator>
            <pubDate>Mon, 29 Apr 2019 16:03:01 GMT</pubDate>
            <atom:updated>2019-04-29T17:18:28.438Z</atom:updated>
            <content:encoded><![CDATA[<p>Tl;dr: We’re still <a href="http://test.ai/">test.ai</a>. We’re just updating our logo, so keep calm and test on.</p><h4><strong>The Change</strong></h4><p>As you may have noticed, we look different. And you might be asking why the change? Well, the number one reason is that the previous logo looked generic. Yep, we said it.</p><p>If you do a google image search for ‘AI’, you’ll see this neural network pattern repeated over and over.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7zEUSCk8l0ibVwmpufr0qQ.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*4hJX9jajXtpKLkcpVmQJEw.jpeg" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nf0FvEI6yjQQX0vvQvjNUw.jpeg" /></figure><p>It also happens that, at this moment, every new Linkedin profile has a background image of a similar pattern. While many of these images and variations do look pretty cool, we felt like we could do better.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*tcdm24Rcs6b1cfsYydoyxA.png" /></figure><p>Even with our old logo, sometimes, people would think we were a scientific lab testing company. Given that the logo could look like a molecule, that was a fair conclusion. Overall, it was fairly bland and gave a misleading idea as to what we do here at test.ai. In case you need a reminder, this was our old one:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BkiYxZjE4C75-PkaKQYCAg.png" /></figure><p>Now this is our new one:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x1Brvejp3V572x-YNRB_dw.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eGXDyvDuZzPqFb9k2pI4Dg.png" /></figure><p>And here’s an animated version of the mark:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/596/1*fVb9vRonkqir6El-I5RYkw.gif" /></figure><p>We’re not going to go into all the minutiae of the design and the reasoning behind it, but what we will say is that we feel this one feels like us. We’re moving testing into its next phase with our automated platform and want to help teams scale test automation.</p><h4><strong>What Next?</strong></h4><p>Our facelift doesn’t end here. Over the coming weeks, you’ll be seeing massive changes to our website. Massive, <em>useful</em> changes. What sort of changes?</p><p>You’ll be able to:</p><ul><li>Browse some of the mobile apps we’ve tested</li><li>Explore common tests run across apps</li><li>View lots and lots of screenshots</li></ul><p>So sign up for our <a href="https://test.ai/">mailing list</a> to get notified when it’s up! You won’t want to miss it.</p><p>— Jason Arbon, CEO at test.ai</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=619b351b4f7b" width="1" height="1" alt=""><hr><p><a href="https://medium.com/testdotai/test-ai-has-a-new-look-619b351b4f7b">Test.ai Has a New Look</a> was originally published in <a href="https://medium.com/testdotai">testdotai</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Black Friday User Experience]]></title>
            <link>https://medium.com/testdotai/the-black-friday-user-experience-aeb0c02172c6?source=rss----4417152e2fa4---4</link>
            <guid isPermaLink="false">https://medium.com/p/aeb0c02172c6</guid>
            <category><![CDATA[test-automation]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <category><![CDATA[retail]]></category>
            <dc:creator><![CDATA[test.ai]]></dc:creator>
            <pubDate>Wed, 21 Nov 2018 19:22:41 GMT</pubDate>
            <atom:updated>2018-11-21T19:22:40.838Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ZWn6KLG89ay6i-ob4ZgHoA.png" /></figure><p>It is that time of year again where people are rushing to get the best deals on <em>Black Friday</em>. The difference this year? Shoppers are not camping out in the parking lot as much as years past — they are grabbing their phone, launching their favorite shopping app, and hoping for a quick and easy experience to do all their holiday shopping.</p><p>So, which are the best apps for Black Friday shopping? I’m going to compare Amazon, Target and Walmart apps across core workflows that most shopping apps have: finding deals, finding gifts and purchasing.</p><p><strong>Finding Deals: Those Black Friday Ads</strong></p><p>If you want the best deals, and fast, you want a shopping experience that quickly gets right to the ads with all the deals. How many clicks does it take to get there? And how long does it take the ad pages of these apps to load? Let’s take a look:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qYAWHl5fXszDr1MAGgRPpw.png" /><figcaption>Amazon: 2 steps from home screen</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*KG9m2IZY6sTvykPnXlUf1g.png" /><figcaption>Target: 7 steps home home screen</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*BMzJoJlleMVzvPIhStG7DQ.png" /><figcaption>Walmart: 3 steps from home screen</figcaption></figure><blockquote><strong>Finding Deals</strong><br>Amazon: 2 steps<br>Target: 7 steps<br>Walmart: 3 steps</blockquote><p>Walmart only makes you tap three times to get to the ads. Target doubles as a workout app and has you tapping seven times to get to the ads. Amazon is the champ and takes just two steps to get to the Black Friday ads. Amazon does has the advantage of having only one ad since they don’t have large box stores with different ads across the country. But that’s a good thing for the user — simplicity.</p><p>When you do finally get to the weekly ads, Target’s deals load almost instantaneously. Walmart’s deals take about a full three seconds to load — that’s a long time in the mobile world. Disappointingly, with all of Amazon’s engineering prowess, and the fact that they only have to deal with a single ad, their ads loaded the slowest at 6.6 seconds . That’s a puzzler. The engineers at Target are doing something smart and making the experience super fast for their Black Friday shoppers… once they get there.</p><blockquote><strong>Loading Deals</strong><br>Amazon: 6.6 seconds<br>Target: Almost instant<br>Walmart: 3.1 seconds</blockquote><p>It is tougher to find the ads in the Target app — you might get frustrated, or not even find them. But, when you do, it loads really fast. Walmart takes users to the ads quickly but then leaves them waiting for a page load. Which is best for Black Friday shopping depends on whether you like to click a lot or wait a lot :-/ . Seems the three teams could learn a lot from each other and hopefully next year the ads in all the apps are instantaneous and just a couple clicks away.</p><blockquote><strong>App Quality Tip: Pre-fetch<br></strong>If you know you will need data, and what data you will need, fetch it before the user asks. Then you can look like a champ when you display it instantly, not just on-demand like an old school web link click. Pull all that data down locally and get it ready before the user asks for it. If you don’t, your app won’t feel buttery. Remember, today’s smart phones are little super computers with lots of storage — leverage it for elegance.</blockquote><p>The best app for viewing ads? Looking at both the number of steps it takes to find them and how long it takes to load them… it’s an anticlimactic three-way tie.</p><p><strong>Finding Gifts: Searching For that Specific Toy</strong></p><p>I have kids, and every year, they somehow know the exact toy they want. However, the toys they want might not always be in an ad. So, how do we find them? We search!</p><p>Let’s see how the search experiences stack up and search for LEGOs across our shopping apps:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*eU1CP0o1HeWimJ-_54jdEA.png" /><figcaption>Searching on Amazon is just a couple steps</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2ZcFCUNYhw8heOHuJ5-WeQ.png" /><figcaption>Searching in Target’s app is a bit more cumbersome</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*XYQecNKbxn92obQI4Vqcqg.png" /><figcaption>Searching in Walmart’s app is easy</figcaption></figure><blockquote><strong>Searching for Toys</strong><br>Amazon: 2 steps<br>Target: 4 steps<br>Walmart: 2 steps</blockquote><p>Target’s search experience is pretty typical. In all, it takes about 4 steps to find a LEGO product page. The Walmart and Amazon search experiences get their users to the products in fewer steps.</p><blockquote><strong>Loading Search Results</strong><br>Amazon: 1.6 seconds<br>Target: 3.7 seconds<br>Walmart: 3.8 seconds</blockquote><p>Amazon’s search results are over twice as fast as Walmart and Target. Twice as fast is a huge deal, and can lead to differences in overall user perception of app quality. Amazon’s users aren’t left staring at the screen waiting on the app.</p><blockquote><strong>App Quality Tip: Cache</strong><br>Cache everything. Well, everything that’s reasonable. Don’t keep pulling your company’s logo down via URL every time the app is launched. Unless you are building a real-time stock quote app, you can probably cache almost all your URL calls. It is too easy to do this right with libraries like SDWebImage. Cache. Cache. Cache.</blockquote><p>The best app for searching for toys? Amazon.</p><p><strong>Purchasing: Signing in and Checking Out</strong></p><p>Now that we’ve found some great deals on some great toys, how do we buy them? You need to be logged in and logging into an app can be frustrating. Let’s see how these apps did:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7UW8HOYmDzdIG1ngeH2BEQ.png" /><figcaption>Signing into Amazon’s app is short and sweet</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*6pgon6bVhdBqH0FJJA8_BQ.png" /><figcaption>Signing into Target’s app bounces users to a web page and pops permission dialogs</figcaption></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*iS7VecxrZoLFUiPUXK1Ing.png" /><figcaption>Signing into Walmart’s app is quick and simple</figcaption></figure><blockquote><strong>Signing In</strong><br>Amazon: 3 steps<br>Target: 9 steps<br>Walmart: 4 steps</blockquote><p>Amazon and Walmart again seem to have better app designs, better flows for users to sign-in. However, did you notice that the Walmart sign-in button can get stuck behind the keyboard? Target’s users not only need more clicks to sign-in, but also are forced to go through permission dialogs after they sign-in whether they like it or not. Target users also have to load a web browser in the process — great apps have a native login user interface, or at least have the browser hidden from the user (hybrid app).</p><blockquote><strong>Sign-in time (after username and password)</strong><br>Amazon: 5.3 seconds<br>Target: 8.7 seconds<br>Walmart: 5 seconds</blockquote><p>While both Amazon and Walmart clock in at around five seconds, five seconds is still a long time for users to wait for the app to sign-in. Try counting to five slowly right now. Now imagine the millions of people using that app who will be wasting their time signing in this Black Friday. Worse, these are customers that have a full shopping cart and are trying to buy toys. These apps are relying on people to be patient and free of distraction during those five seconds.</p><blockquote><strong>App Quality Tip: Go Native</strong><br>It is very tempting to just load your responsive website into a hybrid app. These types of apps can be built in just days. It is also tempting to just point users to a mobile web page. Your users will prefer the native version if you can afford to build a nice one. Native, especially on iOS, is just plain faster, looks better, and has lots of transitions and effects that you just can’t get with HTML5 today. If you want an elegant app, it has to be native.</blockquote><p>The best app for signing in? Amazon by a keyboard.</p><p><strong>Our Winner: Amazon</strong></p><p>From our analysis above, Amazon is the best of the three apps for Black Friday shoppers. And let’s be honest with ourselves — these stores are all selling the same basic toys, all at about the same price. The major difference is the shopping experience.</p><p>Our shopping experiences, along with most of our other experiences, are moving to mobile. Those experiences have to be awesome to remain competitive. It seems that apps are getting better year by year, but there is still plenty for app designers, developers, and testers to learn between now and the next Black Friday come 2019.</p><p>— Jason Arbon, CEO at test.ai</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=aeb0c02172c6" width="1" height="1" alt=""><hr><p><a href="https://medium.com/testdotai/the-black-friday-user-experience-aeb0c02172c6">The Black Friday User Experience</a> was originally published in <a href="https://medium.com/testdotai">testdotai</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[AI Beat 70 Testers — and You!]]></title>
            <link>https://medium.com/testdotai/ai-beat-70-testers-and-you-c4d14e8f8007?source=rss----4417152e2fa4---4</link>
            <guid isPermaLink="false">https://medium.com/p/c4d14e8f8007</guid>
            <category><![CDATA[testing]]></category>
            <category><![CDATA[app-development]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[mobile-app-development]]></category>
            <dc:creator><![CDATA[test.ai]]></dc:creator>
            <pubDate>Wed, 31 Oct 2018 16:47:30 GMT</pubDate>
            <atom:updated>2018-10-31T16:47:30.432Z</atom:updated>
            <content:encoded><![CDATA[<h3>AI Beat 70 Testers — and You!</h3><p>AI can beat humans in the games of Chess, Jeopardy, and Go. Recently, I’ve also discovered that AI systems can also beat 70 human testers at their own game — testing. I spend my days coercing machines to execute test cases with techniques such as neural networks, clustering, and reinforcement learning. It is a lot of work, but it is somewhat mechanical and obvious that with AI driving cars around town and playing Atari video games these days, the AI should also be able to execute a test case or two for us.</p><p>Testers continually ask me if their jobs are safe, and I’ve usually said yes for three reasons:</p><ol><li>Human minds are still writing the actual test cases today and the AI just executes it for us</li><li>When I’m asked this question, I’m typically surrounded by 50 to 1000 testers and I do value my life</li><li>There have always been areas of testing that I thought could not be relegated to the machines for a very long time, specifically qualitative assessments of quality</li></ol><p>Qualitative assessments of quality are things such as: How easy is this app to use? Does this app look good? Is this screen trustworthy? If we humans, myself included, are confused by some of these qualitative judgments of quality, then surely our jobs would be safe from the machines if we were to focus on these areas.</p><p><strong>The Challenge</strong></p><p>I co-presented a full-day tutorial on the topic of AI and machine learning for testers. The room had 70 testers in the room. These testers are no ordinary testers, these were professional, technical testers who work in roles where their company can afford to send them off to Disneyland for a week of training. These testers were also confident enough in themselves to brave a full day learning about AI and machine learning algorithms. The room was full of great testers.</p><p><em>Let’s pretend you are in that room with us and see if the AI can beat you too.</em></p><p>I asked the audience, and now I’m asking you, a qualitative testing question:</p><blockquote>Question: If you were looking at screenshots of login pages, how would you tell if the login page is trustworthy or not? How would you rate the page’s trustworthiness?</blockquote><p>Think about it for a moment…</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*nQLsUapW6_Pbjd1nI--aYw.png" /><figcaption>Samples of Untrustworthy and Trustworthy Login Pages</figcaption></figure><p>Take your time and look at some of the example login screens above.</p><p>What do you think?</p><p><strong>The Results: 70 Experts</strong></p><p>The other 70 people in the room took their time to think about this one too. There were no quick answers. A general ‘hmmmph’ sounded in the room. To make folks feel better about drawing a blank, I said that I originally had no idea either. I encouraged folks to keep thinking. A woman in the front row (who’s son works on AI at Google), ventured a guess and said: “foreign languages”. If the login page was in the US app store, but it had non-English words in it, it could be viewed as less trustworthy to a user because they wouldn’t know what the app was telling them. A great start, but only after some 3 minutes of 70 human minds thinking in parallel.</p><p>After a few more moments, a second hand went up, and this gentleman suggested that if the login had a well trusted and recognized brand name or logo, the login page would be more trustworthy. If the login page showed the Google or Microsoft logo, then it is probably more trustworthy than the average app. If the login page had a logo the user had never seen or heard of before, it would be seen as less trustworthy.</p><p>At this point, we burned perhaps 70 * 5 minutes of human compute time, which is about 6 hours of top human testing brain power trying to answer this question. And yet, we only have 2 ideas on how to measure trustworthiness. We also don’t have a method of scoring exactly how trustworthy or untrustworthy a login screen is — just some ideas on what might influence that qualitative judgment of quality.</p><p><strong>The Results: AI</strong></p><p>OK, so now we timeout on the human minds because they can be fickle, get impatient and bored. How the did ‘AI’ do? Previously, at test.ai world headquarters, we had a few non-technical humans simply look at a large number of login screens and rate the trustworthiness of the pages from 1 to 10. Then, we simply trained a neural network with this data. What was the AI’s answer to this question of how to determine the trustworthiness of a login page? We peeked inside the neural network weights and found the AI’s answer to the question was:</p><ol><li>Foreign characters — If the login page has foreign characters or words in it, it is less trustworthy.</li><li>Brand recognition — If the login page has a popular/recognizable brand image, it is more trustworthy.</li><li>Number of elements — If the page had more elements, it was less trustworthy. The fewer elements on the page, the more trustworthy.</li></ol><p>The machines beat the humans! Not only did the AI discover an additional aspect of the application that correlates to trustworthiness (#3), but the AI also gave a precise score of how trustworthy or untrustworthy a login page is on a scale of 1 to 10.</p><p>Interestingly, the trained AI was smarter than any single human that labeled the data. The neural network effectively learned the collective intelligence of all humans that provided the training data. For example, if someone was fluent in Chinese and familiar with Chinese software, they might not have viewed those login pages as untrustworthy. The AI is smarter than any single human.</p><p>The AI was also more reliable than the humans. Remember, we asked 70 great testers this question. Given this sample, a company only has a 2 in 70 (2.8%) chance of having a tester on their staff that could answer this question. Even then, no single human figured out more than one aspect of trustworthiness, where the AI discovered 3 aspects.</p><p>Even more so, the AI reflects how ‘real’ users view trustworthiness. The human testers are trying to emulate that assessment indirectly with empathy — but they are not the actual target user. How much better is it to have the oracle be real-world users versus a tester trying to reverse engineer and guess at end-user qualitative assessments of quality?</p><p>The AI is also a whole lot less expensive, reusable for all app teams, and faster than humans. The trained AI takes about 200 milliseconds to measure trustworthiness.</p><p><strong>AI Can Teach Us Too</strong></p><p>This was a day that AI beat 70 humans (and likely you) at what was commonly viewed as one of the tougher problems in testing — qualitative assessment. It is common for people to dismiss the advance of AI because they know that their particular profession would be difficult for AI to reproduce. But the reality is that AI can often do a better job than humans in many professional fields including software engineering and testing. Testers will soon join the ranks of radiologists, lawyers, and truck drivers who have to deal with the uncomfortable truth that AI might just better than they are at their job.</p><p>AI beating humans in software testing tasks — some of the most difficult tasks, was fascinating. But after some time considering what had happened, this was also the first time an AI had taught me (and you!) something about testing and quality. We both now know that if we have more elements, check boxes, text, icons etc. on a login page, it will likely be viewed by users as less trustworthy. You didn’t know that a few minutes ago. If we hadn’t looked inside the AI’s brain, I’m not sure any human would know that today. Humans are now learning from AI and that might be more significant in the long run than them beating the humans at our own software testing games.</p><p>— Jason Arbon, CEO at test.ai</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4d14e8f8007" width="1" height="1" alt=""><hr><p><a href="https://medium.com/testdotai/ai-beat-70-testers-and-you-c4d14e8f8007">AI Beat 70 Testers — and You!</a> was originally published in <a href="https://medium.com/testdotai">testdotai</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Training Data for App Classifier]]></title>
            <link>https://medium.com/testdotai/training-data-for-app-classifier-f217dc005523?source=rss----4417152e2fa4---4</link>
            <guid isPermaLink="false">https://medium.com/p/f217dc005523</guid>
            <category><![CDATA[app-icon]]></category>
            <category><![CDATA[training-data]]></category>
            <category><![CDATA[appium]]></category>
            <category><![CDATA[classifier]]></category>
            <dc:creator><![CDATA[test.ai]]></dc:creator>
            <pubDate>Wed, 24 Oct 2018 16:59:45 GMT</pubDate>
            <atom:updated>2018-10-24T16:59:44.693Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*x82i7LUsX3p_HGoE" /></figure><p>Test.ai open sourced an AI classifier for Appium at SeleniumConf last week (<a href="https://www.linkedin.com/pulse/adding-ai-appium-step-by-step-guide-android-emulator-alejandro-avella/">setup article</a>). This post will show you how to build the classifier yourself, customize them for your application, and add it to Appium.</p><p>Before you begin, you will first need to install TensorFlow on your machine. TensorFlow is the open source AI framework from Google.</p><pre>pip install --upgrade tensorflow</pre><p><strong>Classify Elements</strong></p><p>Checkout the test.ai repo (<a href="https://github.com/testdotai/classifier-builder">https://github.com/testdotai/classifier-builder</a>) on github, which will include all the necessary components (images, classifier, building script), and <em>cd</em> into the sample_run directory.</p><p>We will run the classifier on a cart:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/60/0*TH6jeN3j4vyeSeBw" /></figure><p>Run the script:</p><pre><em>python run_model.py --image cart.png</em></pre><p>This outputs — <em>Top results: cart — 0.7937941</em></p><p>The 0.7937941 tells you that the AI is 79.38% sure that it is a shopping cart. From our experience, anything above a confidence of 0.3 is usually confident enough.</p><p><strong>Build A Classifier</strong></p><p>Run the following command from the repo root (<em>cd ..)</em> to start the training:</p><pre><em>python retrain.py --image_dir training_images/ --output_graph output/saved_model.pb --output_labels output/saved_model.pbtxt --how_many_training_steps 4000 --learning_rate 0.30 --testing_percentage 25 --validation_percentage 25 --eval_step_interval 50 --train_batch_size 2000 --test_batch_size -1 --validation_batch_size -1 --bottleneck_dir /tmp/bottleneck</em></pre><p><strong>NOTE</strong>: <em>The initial training will take 2+ hours on a modern MacBook Pro.</em></p><p><strong>How to Add to the Data Set</strong></p><p>If you wish to add a new image that isn’t in the data set, such as a “refund_button”. Create a new folder called “refund_button” under training_images/, and fill it with images of refund buttons, followed by re-running the training.</p><pre><em>python retrain.py --image_dir training_images/ --output_graph output/saved_model.pb --output_labels output/saved_model.pbtxt --how_many_training_steps 4000 --learning_rate 0.30 --testing_percentage 25 --validation_percentage 25 --eval_step_interval 50 --train_batch_size 2000 --test_batch_size -1 --validation_batch_size -1 --bottleneck_dir /tmp/bottleneck</em></pre><p><strong>Updating Appium to use your Data</strong></p><p>Now that you have an updated version of the classifier, it’s time to add it to Appium.</p><p>Convert the classifier to tensorflowjs. Install tensorflowjs.</p><pre>pip install tensorflowjs</pre><p>Run the conversion:</p><pre><em>tensorflowjs_converter --input_format=tf_frozen_model --output_node_names=final_result  output/saved_model.pb web_model</em></pre><p>This creates a web_model folder. Copy the contents of this folder to the npm dependency install for appium-classifier-plugin under the model folder (/usr/local/lib/node_modules/appium/node_modules/test-ai-classifier/model/).</p><p><strong>Help Improve the Data Set</strong></p><p>We will regularly update the data set and encourage community involvement to improve it. Feel free to send any pull requests with additional images/classes to add to the data set. For more information on how to do a github pull request, check out: <a href="https://help.github.com/articles/creating-a-pull-request-from-a-fork/">https://help.github.com/articles/creating-a-pull-request-from-a-fork</a>.</p><p>Finally, there is a chance that tests won’t break on every random build…</p><p>— Chris Navrides, Engineer at test.ai</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f217dc005523" width="1" height="1" alt=""><hr><p><a href="https://medium.com/testdotai/training-data-for-app-classifier-f217dc005523">Training Data for App Classifier</a> was originally published in <a href="https://medium.com/testdotai">testdotai</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Adding AI to Appium]]></title>
            <link>https://medium.com/testdotai/adding-ai-to-appium-f8db38ea4fac?source=rss----4417152e2fa4---4</link>
            <guid isPermaLink="false">https://medium.com/p/f8db38ea4fac</guid>
            <category><![CDATA[software-testing]]></category>
            <category><![CDATA[appium]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <dc:creator><![CDATA[test.ai]]></dc:creator>
            <pubDate>Thu, 18 Oct 2018 08:01:02 GMT</pubDate>
            <atom:updated>2018-10-18T15:29:10.142Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*5hoLTS-eFqBaztWJcr03sg.png" /></figure><p>With one API call, you can add the power of AI to your mobile test automation. The team at test.ai has teamed up with <a href="https://www.linkedin.com/in/jlipps">Jonathan Lipps</a>, the lead contributor of Appium and founder of <a href="https://cloudgrey.io/">Cloud Grey</a>, to add a bit of AI to Appium. The AI finds common elements in mobile apps such as search text boxes, login buttons, etc., so test developers don’t have to worry about all those magic IDs, CSS, or XPaths. Just tell the AI what you want it to find and it will find it for you on the page — even if the element changes color, text, location, or position in the DOM. With AI, tests will be a little quicker to create and will break a little less often. Welcome to the world of automation powered by AI!</p><p><strong>How to Get Started<br></strong>How do you leverage the power of this AI brain in your code?</p><ol><li>Simply update your Appium project to the latest revision (see Jonathan Lipp’s <a href="https://appiumpro.com/editions/39">Appium Pro article</a> for the details)</li><li>Then, find elements using a new custom AI-search strategy such as:</li></ol><blockquote>driver.elementByCustom(‘ai:cart’);<br>// <em>This code asks the AI to find a shopping cart image on the screen for you</em></blockquote><p>I really can’t imagine it getting easier than this to add AI to your automation project. In fact, it is faster and simpler than the traditional methods of finding IDs, CSS or XPaths and using these more complicated search strategies.</p><p><strong>AI For the Planet<br></strong>Open source is key. No team should have to find the XPath or CSS Selector of an element or beg a developer to add a magic ID for them to use in their test code. No team should have to re-invent basic AI element classifiers or re-label 100,000+ images either — what a waste of humanity to duplicate that work. Therefore, the classifier is open source. Many vendors consider this type of IP their magic sauce, but that means that most test automation engineers can’t afford it, or don’t want to integrate it into their own code base. Open source means this tech is awesomely universally accessible to all.</p><p>Extensibility is key. This is the ‘hello world’ of bringing AI to element finding. Jonathan made sure this was a pluggable system, so any classifier can be used, or even other element search algorithms can be easily shared and plugged directly into Appium. Test.ai just open-sourced the <a href="https://github.com/testdotai/appium-classifier-plugin">default/reference implementation</a> and donated to the community in the interest of sharing the power of AI with every test developer on the planet. Our mission at test.ai is to test the world’s apps. What better way is there than to help every test developer with the basics of finding elements inside of their own apps?</p><p>Customization is key. The testing community can improve the AI. Anyone can add new training data, alternative training methods, more rigorous relevance testing, or new labels. The AI is the property of the community, and we hope to help bootstrap every test team on the planet with a foundation of AI for their own projects. The test.ai team has shared all the <a href="https://www.kaggle.com/testdotai/common-mobile-web-app-icons">training data on Kaggle</a>, so the world can fork the data, clean it up, add it to their proprietary test frameworks, or compete to improve these classifiers. Crowdsourced, open data for AI testing systems? It is a new world.</p><p>Reuse is key. The AI can be forked and/or re-used in other open source and proprietary frameworks. The goal at test.ai is to spread the usage of AI in all aspects of testing, in the interest of faster and smarter test automation, and ultimately better software in the world. The neural networks are based on the open source <a href="https://github.com/tensorflow/tensorflow">TensorFlow</a> framework from Google. These models can be run in the cloud, locally, on mobile devices, or in a project not even thought of yet.</p><p><strong>Become an AI Test Automation Engineer<br></strong>Whether you are an AI expert, test automation geek, or just getting started with AI and testing, you can bring AI into your team and your project today. A free, open source, and single API call is all it takes. You can be the hero that brings a bit of AI to your engineering team. You can even contribute to this transformation in testing by helping add new training data or add a similar call to your favorite test framework — be an AI test automation engineer today. Ultimately, it will take our community to bring the power of AI to our entire field.</p><p>Seeing is believing and Jonathan Lipps created the first intro video demonstrating this working on both Android and iOS. What about web? It works there too, but it’s far less tested.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Ffast.wistia.net%2Fembed%2Fiframe%2Fm6rl8zg77h&amp;url=https%3A%2F%2Ftestai.wistia.com%2Fmedias%2Fm6rl8zg77h&amp;image=https%3A%2F%2Fembed-ssl.wistia.com%2Fdeliveries%2F9126e36abdf7c91bfb3954d20af327165847fffb.jpg%3Fimage_crop_resized%3D960x540&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;schema=wistia" width="960" height="540" frameborder="0" scrolling="no"><a href="https://medium.com/media/267c558717b1c421cac0409da6fb60c2/href">https://medium.com/media/267c558717b1c421cac0409da6fb60c2/href</a></iframe><p>This is a hello world of real AI integrating with test automation tools to make our lives just a little easier, and hopefully more fun.</p><p>By the way, it is awesome working with Jonathan Lipps — truly today’s top expert in mobile test automation. Thanks to the team of machine learning and integration engineers at test.ai for the willingness and bravery to open source something you have poured so much energy and talent into the past year. And, special thanks to our investors who thought this was a great idea when I brought it up.</p><p>“That’s one small step for AI, one giant leap for test automation” :)</p><p>— Jason Arbon CEO, test.ai</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=f8db38ea4fac" width="1" height="1" alt=""><hr><p><a href="https://medium.com/testdotai/adding-ai-to-appium-f8db38ea4fac">Adding AI to Appium</a> was originally published in <a href="https://medium.com/testdotai">testdotai</a> on Medium, where people are continuing the conversation by highlighting and responding to this story.</p>]]></content:encoded>
        </item>
    </channel>
</rss>