<?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 Elodie NGANTCHOU KEMADJOU on Medium]]></title>
        <description><![CDATA[Stories by Elodie NGANTCHOU KEMADJOU on Medium]]></description>
        <link>https://medium.com/@elodiekema2009?source=rss-147ad6d8c05f------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*uKvBrLSE5ABKxU1xduO4-w.jpeg</url>
            <title>Stories by Elodie NGANTCHOU KEMADJOU on Medium</title>
            <link>https://medium.com/@elodiekema2009?source=rss-147ad6d8c05f------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Fri, 20 Nov 2026 23:31:11 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@elodiekema2009/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[Project course two class weather classification]]></title>
            <link>https://medium.com/@elodiekema2009/project-course-two-class-weather-classification-9dc1e8a2c3ef?source=rss-147ad6d8c05f------2</link>
            <guid isPermaLink="false">https://medium.com/p/9dc1e8a2c3ef</guid>
            <dc:creator><![CDATA[Elodie NGANTCHOU KEMADJOU]]></dc:creator>
            <pubDate>Wed, 01 Jul 2020 11:47:14 GMT</pubDate>
            <atom:updated>2020-07-03T11:37:03.385Z</atom:updated>
            <content:encoded><![CDATA[<h3>Two Class Weather Classification</h3><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*2brxm4wEFf6lg9Wx" /><figcaption>Photo by <a href="https://unsplash.com/@jplenio?utm_source=medium&amp;utm_medium=referral">Johannes Plenio</a> on <a href="https://unsplash.com/?utm_source=medium&amp;utm_medium=referral">Unsplash</a></figcaption></figure><p>A good amount of the code I will reference was originally written by the course’s instructor, <a href="https://medium.com/u/ca6fe690e934?source=post_page-----528794240fa1----------------------">Aakash N S</a>, so big thanks to him! This is intended to be a basic overview of some of the things I have learned over a 6-week course and there are a LOT of very complicated concepts that I will just gloss over, but I have tried my best to include relevant links to docs and further reading where it felt like a good idea. I encourage you to click them — these concepts are HUGE and there is so much more to learn.</p><blockquote>For the course project, i will pick any dataset of my choice and apply the concepts learned in this course to train deep learning models end-to-end with PyTorch, experimenting with different hyperparameters &amp; metrics.</blockquote><blockquote>For this project, i use Convolutional Neural Network to Classify two class weather</blockquote><h3>Let’s start to build a model</h3><p>The notebook this is all based on can be found <a href="https://jovian.ml/kemadjou-elodie/project-course-two-class-weather-classification">here</a>, if you’d rather just see all the code without my commentary.)</p><p>The first thing we will want to do is import a bunch of handy libraries:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D4&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D4&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/becd51d1d00215f5350813be0b5cdc74/href">https://medium.com/media/becd51d1d00215f5350813be0b5cdc74/href</a></iframe><p>With our modules imported, and our Kaggle dataset loaded (<a href="https://www.kaggle.com/polavr/twoclass-weather-classification">here</a>), we can get started creating our data loaders.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D6&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D6&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/090cf3e1a3a3e9063eb58ee5dd1beae8/href">https://medium.com/media/090cf3e1a3a3e9063eb58ee5dd1beae8/href</a></iframe><p>Before we go further, we should talk a bit about the three “kinds” of data we will be interacting with; training, test, and validation.</p><h3>Training Data</h3><p>The training data is classified data that we will be showing to our model in batches (more on that soon). Everything that the model will “learn” about bird images will be from this data. After each batch of data is run through the model, we will use gradient descent to adjust the model’s weights and biases.</p><h3>Validation Data</h3><p>At the end of each training epoch, we will use another set of classified data to test how well our model is performing. Obviously, the model would be able to easily identify any image we supplied it during the training phase since it would have seen it already, so having a separate subset of data to use for validation is required.</p><h3>Test Data</h3><p>Once we have finished training our model over several epochs, we will use the test set to report final accuracy. This data can also be used to compare the accuracies of different versions of the model.</p><p>Now that we understand the three types of data we will be working with, let’s load all the data we will be working with!</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D8&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D8&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/0deb84527ccbacc2e5747b6bd3cde865/href">https://medium.com/media/0deb84527ccbacc2e5747b6bd3cde865/href</a></iframe><p>We can look at one image (and its label) in the test set to make sure everything is looking good, and also see a grid of images that represents a batch from the dataloader.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D9&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D9&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/d9d227d77646bf634e80fc2ed68abaa1/href">https://medium.com/media/d9d227d77646bf634e80fc2ed68abaa1/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D10&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D10&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/967f81da239e9cdbe85ddd76a38ac199/href">https://medium.com/media/967f81da239e9cdbe85ddd76a38ac199/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D11&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D11&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/4051e740dbc4d678f9b34e5d3bc913a4/href">https://medium.com/media/4051e740dbc4d678f9b34e5d3bc913a4/href</a></iframe><p>We use the ImageClassificationBase that we covered earlier, but we tell our network to use a pre-trained — torchvision.models.resnet34. We add a new last layer, which takes the inputs from the previous layer of the network and outputs the number of species we are classifying, as previously stated. After we move our model to the GPU</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D13&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D13&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/e14caa9ec863a0accf970c6e72af936d/href">https://medium.com/media/e14caa9ec863a0accf970c6e72af936d/href</a></iframe><p>We will call the fit function to train our model, which accepts a number of epochs, a learning rate, the training and validation dataloaders, and an optimization function. We will be using Stochastic Gradient Descent or SGD here.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D20&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D20&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/3e3f6dc22518705322abfc0072c4c3bd/href">https://medium.com/media/3e3f6dc22518705322abfc0072c4c3bd/href</a></iframe><p>The fit function does a few things. First, it initializes a history list where we will keep track of how the model is performing across epochs. Next, we initialize the optimizer with the model’s parameters and learning rate. We could just call torch.optim.SGD() here, but it is handy to abstract this so we can experiment with different optimization functions (which we will do when we instantiate our model). Next, we get into the main <a href="https://ml-cheatsheet.readthedocs.io/en/latest/gradient_descent.html">gradient descent</a> epoch loop, which is basically:</p><ul><li>Call the model’s training_step method for each batch in our training loader. This makes predictions and returns the loss.</li><li>Get the loss, and then call loss.backward(), which returns the <a href="https://en.wikipedia.org/wiki/Gradient">gradient</a> of the loss.</li><li>Call the optimizer function, which updates the model’s weights and biases based on the loss’s gradient.</li><li>Reset the gradients to zero before we process the next batch</li><li>After this has been done for each batch, run tests against the validation data to determine the model’s accuracy.</li><li>Add the accuracy/loss data to our list for reporting</li></ul><p>And that is basically it! Not very much code to represent the extremely powerful process of gradient descent, which allows our model to “learn”.</p><p>The evaluate method() will describe the model&#39;s performance, and is a handy way to keep track of how it is doing. This is not going to be impressive (as a random sampling of predictions showed us), but let’s evaluate the model before we begin training:</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D21&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D21&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/b022a233e30f7c9433983219a6d2f5b0/href">https://medium.com/media/b022a233e30f7c9433983219a6d2f5b0/href</a></iframe><p>The untrained model’s accuracy is about 0.6. That is really bad, but it’s about what you’d expect at this point.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D23&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D23&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/885ab392a565d492bffb08356855d81e/href">https://medium.com/media/885ab392a565d492bffb08356855d81e/href</a></iframe><p>We will train for 10 epochs, use the Adam optimization function, and have a learning rate of 0.001. These properties are the “hyperparameters” of our model, and changing them can make huge differences in training and performance. You should try changing them and see how it goes!</p><p>Before we train — one more very important thing! Since we are using a pre-trained ResNet, we will want to freeze the model so we only train the last layer (which we added).</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D25&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D25&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/74d3d3bac530c2fc421fcf3850add142/href">https://medium.com/media/74d3d3bac530c2fc421fcf3850add142/href</a></iframe><p>Our model has learned quite a bit! Plotting the training/validation loss and accuracy per epoch shows that it got smart rather quickly. You might notice that the accuracy actually decreased in some epochs — this is likely a result of overfitting, which is a big problem in machine learning, but is unfortunately outside the scope of this post.</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D27&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D27&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/bae8dde462a4c68bcc7a7b78982881bd/href">https://medium.com/media/bae8dde462a4c68bcc7a7b78982881bd/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D33&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D33&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/51cef644b4007a15af3e74caf78ebb38/href">https://medium.com/media/51cef644b4007a15af3e74caf78ebb38/href</a></iframe><p>It also claims to be around 82% accurate — let’s run some tests using the test images we set aside earlier and see how it performs now (I’m just showing the outputs for brevity)</p><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D28&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D28&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/50f0725da64b059fd863efc5cebf629f/href">https://medium.com/media/50f0725da64b059fd863efc5cebf629f/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D29&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D29&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/d7bbbc824803dd2545f9bf12bba95f08/href">https://medium.com/media/d7bbbc824803dd2545f9bf12bba95f08/href</a></iframe><p>We have come a long way, but there are a couple of tricks we can use to improve our accuracy. I’m going to start over with a fresh model for this test, but for brevity, I am only going to show the bits of code that have changed.</p><h3>Data Augmentation</h3><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D39&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D39&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/fcbaad3fb4656810f39f927538e48cd6/href">https://medium.com/media/fcbaad3fb4656810f39f927538e48cd6/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D40&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D40&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/6500a162e1011f6f65a4bbb1a95be933/href">https://medium.com/media/6500a162e1011f6f65a4bbb1a95be933/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D45&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D45&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/cc1fdd1f2e141df5ea4db00e4d5a2704/href">https://medium.com/media/cc1fdd1f2e141df5ea4db00e4d5a2704/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D46%26hideInput%3Dtrue&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D46%26hideInput%3Dtrue&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/b8387bd55872c5b4fcba626e4ec9ffc2/href">https://medium.com/media/b8387bd55872c5b4fcba626e4ec9ffc2/href</a></iframe><iframe src="https://cdn.embedly.com/widgets/media.html?src=https%3A%2F%2Fjovian.ml%2Fembed%3Furl%3Dhttps%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D47%26hideInput%3Dtrue&amp;dntp=1&amp;display_name=Jovian&amp;url=https%3A%2F%2Fjovian.ml%2Fkemadjou-elodie%2Fproject-course-two-class-weather-classification%2Fv%2F5%26cellId%3D47%26hideInput%3Dtrue&amp;image=https%3A%2F%2Fstorage.googleapis.com%2Fjvn%2Fassets%2Fkemadjou-elodie%2Fprofile_images%2F8ea6717e3b554c5dabefa83b2944723f%2Fmedium.jpg&amp;key=a19fcc184b9711e1b4764040d3dc5c07&amp;type=text%2Fhtml&amp;scroll=auto&amp;schema=jovian" width="800" height="200" frameborder="0" scrolling="no"><a href="https://medium.com/media/f79de6d42833bb90251f09fa69bc2750/href">https://medium.com/media/f79de6d42833bb90251f09fa69bc2750/href</a></iframe><p>I hope you have learned a bit from this and are inspired to tackle some machine learning problems. This post barely scratches the surface of most of these topics, so if you are interested you should definitely check out the notebook that I referenced here and use some of the code to try to build your own image classification model. If you really want to take the plunge, check out the course materials over at <a href="https://jovian.ml/forum/c/pytorch-zero-to-gans/18">zero to gans</a>. Happy training!</p><p>Thanks to the course’s instructor, <a href="https://medium.com/u/ca6fe690e934?source=post_page-----528794240fa1----------------------">Aakash N S</a> !!!</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=9dc1e8a2c3ef" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>