title
Image Classification using CNN | Deep Learning Tutorial | Machine Learning Project 9 | Edureka

description
🔥 Edureka Machine Learning Certification training (Use Code: YOUTUBE20) : https://www.edureka.co/masters-program/machine-learning-engineer-training This Edureka video on 'Image Classification using CNN' will give you an overview of Image Classification using Machine Learning and will help you understand various important concepts that concern Image Classification with ML. Following pointers are covered in this Image Classification using CNN: 1) Introduction 2) Tools and Frameworks 3) Project ------------------------------------ 🔹Checkout Edureka's Machine Learning Project playlist: https://bit.ly/3ij9Uw7 🔹Checkout Edureka's Machine Learning Python Tutorial playlist: https://bit.ly/3szLTCO 🔹Checkout Edureka's Machine Learning R Tutorial Playlist: https://bit.ly/3duYGlF 🔹Checkout Edureka's Machine Learning Tutorial Blog Series: https://bit.ly/2PX5lIp 🔴Subscribe to our channel to get video updates. Hit the subscribe button above: https://goo.gl/6ohpTV Twitter: https://twitter.com/edurekain LinkedIn: https://www.linkedin.com/company/edureka Instagram: https://www.instagram.com/edureka_learning/ Facebook: https://www.facebook.com/edurekaIN/ SlideShare: https://www.slideshare.net/EdurekaIN Castbox: https://castbox.fm/networks/505?country=in Meetup: https://www.meetup.com/edureka/ ---------Edureka Machine Learning Projects--------- 🔵 Plant Leaf Disease Detection with GUI: https://bit.ly/36Y6l8g 🔵 House Price Prediction using ML: https://bit.ly/3i0VKzJ 🔵 Emoji Prediction using LSTM: https://bit.ly/2TDuPjR 🔵 Color old photographs using Autoencoders: https://bit.ly/3BQg7r9 🔵 Handwritten Digit Recognition on MNIST dataset: https://bit.ly/3zTCxGf 🔵 Generate Images Using DC-Gan's: https://bit.ly/2TPtYwC 🔵 Building Document Scanner Using OpenCV: https://bit.ly/3lqequN 🔵 Cartoon Effect on Image using OpenCV: https://bit.ly/3rV9rTR ---------𝐄𝐝𝐮𝐫𝐞𝐤𝐚 𝐎𝐧𝐥𝐢𝐧𝐞 𝐓𝐫𝐚𝐢𝐧𝐢𝐧𝐠 𝐚𝐧𝐝 𝐂𝐞𝐫𝐭𝐢𝐟𝐢𝐜𝐚𝐭𝐢𝐨𝐧--------- 🔵 Data Science Online Training: https://bit.ly/2NCT239 🟣 Python Online Training: https://bit.ly/2CQYGN7 🔵 AWS Online Training: https://bit.ly/2ZnbW3s 🟣 RPA Online Training: https://bit.ly/2Zd0ac0 ---------𝐄𝐝𝐮𝐫𝐞𝐤𝐚 𝐌𝐚𝐬𝐭𝐞𝐫𝐬 𝐏𝐫𝐨𝐠𝐫𝐚𝐦𝐬--------- 🟣Machine Learning Engineer Masters Program: https://bit.ly/388NXJi 🟣Cloud Architect Masters Program: https://bit.ly/3i9z0eJ 🔵Data Scientist Masters Program: https://bit.ly/2YHaolS 🟣Big Data Architect Masters Program: https://bit.ly/31qrOVv 🔵Business Intelligence Masters Program: https://bit.ly/2BPLtn2 -----------------𝐄𝐝𝐮𝐫𝐞𝐤𝐚 𝐏GD 𝐂𝐨𝐮𝐫𝐬𝐞--------------- 🔵Artificial and Machine Learning PGD: https://bit.ly/2Ziy7b1 #edureka #edurekamachinelearning #machinelearning #ImageClassificationusingCNN #machinelearningproject #machinelearningtutorial #edurekatraining -------------------------------------------------------------------- About the Course : Edureka’s Machine Learning Course using Python is designed to make you grasp the concepts of Machine Learning. The Machine Learning training will provide a deep understanding of Machine Learning and its mechanism. As a Data Scientist, you will be learning the importance of Machine Learning and its implementation in the python programming language. Furthermore, you will be taught Reinforcement Learning which in turn is an important aspect of Artificial Intelligence. You will be able to automate real life scenarios using Machine Learning Algorithms. Towards the end of the course we will be discussing various practical use cases of Machine Learning in python programming language to enhance your learning experience. -------------------------------------- Why Learn Machine Learning with Python? Data Science is a set of techniques that enables the computers to learn the desired behavior from data without explicitly being programmed. It employs techniques and theories drawn from many fields within the broad areas of mathematics, statistics, information science, and computer science. This course exposes you to different classes of machine learning algorithms like supervised, unsupervised and reinforcement algorithms. This course imparts you the necessary skills like data pre-processing, dimensional reduction, model evaluation and also exposes you to different machine learning algorithms like regression, clustering, decision trees, random forest, Naive Bayes and Q-Learning. -------------------------------------------------------------------- Got a question on the topic? Please share it in the comment section below and our experts will answer it for you. For more information, Please write back to us at sales@edureka.co or call us at IND: 9606058406 / US: 18338555775 (toll free).

detail
{'title': 'Image Classification using CNN | Deep Learning Tutorial | Machine Learning Project 9 | Edureka', 'heatmap': [{'end': 266.766, 'start': 208.104, 'weight': 0.903}, {'end': 1137.57, 'start': 1083.519, 'weight': 0.733}, {'end': 1691.048, 'start': 1608.485, 'weight': 0.875}], 'summary': 'This tutorial covers visualizing gradients with feature maps, training deep learning models using tensorflow and keras with gpu, building an image classification model, and visualizing cnn using grad cam for model interpretability.', 'chapters': [{'end': 37.135, 'segs': [{'end': 37.135, 'src': 'embed', 'start': 7.369, 'weight': 0, 'content': [{'end': 15.156, 'text': "Hi all this is Junaid here from Edureka and I welcome you all to this interesting session where we're going to see how we can visualize gradients using feature map.", 'start': 7.369, 'duration': 7.787}, {'end': 18.659, 'text': "So without any further delay, let me quickly walk you through today's agenda.", 'start': 15.516, 'duration': 3.143}, {'end': 23.943, 'text': "First off, we're going to start this session by understanding the problem statement and the workflow that goes along with it.", 'start': 19.019, 'duration': 4.924}, {'end': 30.269, 'text': "following that, I'll be discussing various tools and frameworks that we are going to use now, and then we'll finally jump into the project over here.", 'start': 23.943, 'duration': 6.326}, {'end': 37.135, 'text': 'Okay guys, so before we begin, do consider subscribing to our YouTube channel and hit the bell icon to stay updated on training Technologies.', 'start': 30.809, 'duration': 6.326}], 'summary': 'Junaid from edureka will discuss visualizing gradients using feature map and relevant tools in this session.', 'duration': 29.766, 'max_score': 7.369, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk7369.jpg'}], 'start': 7.369, 'title': 'Visualizing gradients with feature map', 'summary': 'Covers the process of visualizing gradients using feature maps, starting with understanding the problem statement and workflow, discussing tools and frameworks, and finally jumping into the project. it also encourages subscribing to the youtube channel for staying updated on training technologies.', 'chapters': [{'end': 37.135, 'start': 7.369, 'title': 'Visualizing gradients with feature map', 'summary': 'Covers the process of visualizing gradients using feature maps, starting with understanding the problem statement and workflow, followed by discussing tools and frameworks, and finally jumping into the project. the session also encourages subscribing to the youtube channel for staying updated on training technologies.', 'duration': 29.766, 'highlights': ['The session covers visualizing gradients using feature maps, starting with understanding the problem statement and workflow, followed by discussing tools and frameworks, and finally jumping into the project.', 'Encourages subscribing to the YouTube channel for staying updated on training technologies.']}], 'duration': 29.766, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk7369.jpg', 'highlights': ['The session covers visualizing gradients using feature maps, starting with understanding the problem statement and workflow, discussing tools and frameworks, and finally jumping into the project.', 'Encourages subscribing to the YouTube channel for staying updated on training technologies.']}, {'end': 1131.448, 'segs': [{'end': 101.521, 'src': 'embed', 'start': 76.016, 'weight': 1, 'content': [{'end': 83.867, 'text': 'You see when I train this model my data set over here is going to have a 25, 000 images of a cat and similarly it will be 25, 000 images of a dog.', 'start': 76.016, 'duration': 7.851}, {'end': 88.813, 'text': 'So our model will train in different, different variations, but you know when I try to predict this.', 'start': 84.267, 'duration': 4.546}, {'end': 92.258, 'text': "obviously there is going to be a feature map, or let's say, gradient, that I have,", 'start': 88.813, 'duration': 3.445}, {'end': 98, 'text': 'And what this feature map does is wherever the probability or the pattern that represents a cat are high.', 'start': 92.558, 'duration': 5.442}, {'end': 101.521, 'text': 'We will use the feature map or the probability value or that is going to be high.', 'start': 98.16, 'duration': 3.361}], 'summary': 'Training model with 25,000 cat and dog images for accurate prediction using feature map.', 'duration': 25.505, 'max_score': 76.016, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk76016.jpg'}, {'end': 179.347, 'src': 'embed', 'start': 152.396, 'weight': 0, 'content': [{'end': 158.737, 'text': "What we'll do is we'll use some pre-trained model and let's take something like inception model or VGG 19,", 'start': 152.396, 'duration': 6.341}, {'end': 163.599, 'text': "and finally we'll have we're up trainer model and then we'll try to test this up, like how well our model is working.", 'start': 158.737, 'duration': 4.862}, {'end': 164.9, 'text': 'And then the next stage.', 'start': 164.099, 'duration': 0.801}, {'end': 171.663, 'text': "what we're going to do is we're going to use a grad cam method to predict or to get a feature map, and this would be in the form of a heat map.", 'start': 164.9, 'duration': 6.763}, {'end': 179.347, 'text': "and finally, in the last stage, what we'll do is we'll take our original image and then we will paste or will reimburse the heat map on top of this.", 'start': 171.663, 'duration': 7.684}], 'summary': "Using pre-trained models, we'll train and test our model, then apply grad cam method for feature mapping.", 'duration': 26.951, 'max_score': 152.396, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk152396.jpg'}, {'end': 266.766, 'src': 'heatmap', 'start': 208.104, 'weight': 0.903, 'content': [{'end': 213.407, 'text': 'All right guys, so let us know, quickly move to our code editor and see how we can implement this interesting project.', 'start': 208.104, 'duration': 5.303}, {'end': 215.589, 'text': 'as you can see, we have come to a code editor here.', 'start': 213.407, 'duration': 2.182}, {'end': 217.19, 'text': 'Let me change our runtime.', 'start': 215.989, 'duration': 1.201}, {'end': 218.511, 'text': "Usually it's in none.", 'start': 217.39, 'duration': 1.121}, {'end': 220.232, 'text': "So we'll change this to a GPU one.", 'start': 218.571, 'duration': 1.661}, {'end': 221.879, 'text': "and we'll save this.", 'start': 221.159, 'duration': 0.72}, {'end': 226.941, 'text': "So now what we're going to do is as you can see I've already imported a couple of lines of code.", 'start': 222.379, 'duration': 4.562}, {'end': 228.982, 'text': "So first off I'm going to ignore the warning.", 'start': 227.381, 'duration': 1.601}, {'end': 230.222, 'text': "So this is what I'm going to do here.", 'start': 229.022, 'duration': 1.2}, {'end': 234.383, 'text': 'And then these are the basic modules usually used for you know, handling the file.', 'start': 230.782, 'duration': 3.601}, {'end': 235.384, 'text': 'So OS.', 'start': 234.763, 'duration': 0.621}, {'end': 237.384, 'text': 'with OS we can deal with path.', 'start': 235.384, 'duration': 2}, {'end': 242.186, 'text': 'shuttle basically is used to copy the file from one folder to the other and Glob over here is.', 'start': 237.384, 'duration': 4.802}, {'end': 244.006, 'text': 'we usually use Glob for pattern matching.', 'start': 242.186, 'duration': 1.82}, {'end': 245.907, 'text': 'Let me quickly execute this up.', 'start': 244.446, 'duration': 1.461}, {'end': 251.472, 'text': "and we'll go ahead and this is where I have uploaded my data set in this particular Dropbox.", 'start': 246.847, 'duration': 4.625}, {'end': 255.956, 'text': "And once I execute this you will have a zip file and we'll use this part to unzip our files.", 'start': 251.632, 'duration': 4.324}, {'end': 257.857, 'text': 'So let me execute this as well.', 'start': 256.476, 'duration': 1.381}, {'end': 259.478, 'text': 'This will take a moment or two.', 'start': 258.218, 'duration': 1.26}, {'end': 261.1, 'text': "So let's wait for some time.", 'start': 259.978, 'duration': 1.122}, {'end': 266.766, 'text': 'As you can see here, we have successfully imported our data.', 'start': 263.883, 'duration': 2.883}], 'summary': 'Implementing a project with gpu runtime, importing and processing data successfully.', 'duration': 58.662, 'max_score': 208.104, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk208104.jpg'}, {'end': 255.956, 'src': 'embed', 'start': 227.381, 'weight': 2, 'content': [{'end': 228.982, 'text': "So first off I'm going to ignore the warning.", 'start': 227.381, 'duration': 1.601}, {'end': 230.222, 'text': "So this is what I'm going to do here.", 'start': 229.022, 'duration': 1.2}, {'end': 234.383, 'text': 'And then these are the basic modules usually used for you know, handling the file.', 'start': 230.782, 'duration': 3.601}, {'end': 235.384, 'text': 'So OS.', 'start': 234.763, 'duration': 0.621}, {'end': 237.384, 'text': 'with OS we can deal with path.', 'start': 235.384, 'duration': 2}, {'end': 242.186, 'text': 'shuttle basically is used to copy the file from one folder to the other and Glob over here is.', 'start': 237.384, 'duration': 4.802}, {'end': 244.006, 'text': 'we usually use Glob for pattern matching.', 'start': 242.186, 'duration': 1.82}, {'end': 245.907, 'text': 'Let me quickly execute this up.', 'start': 244.446, 'duration': 1.461}, {'end': 251.472, 'text': "and we'll go ahead and this is where I have uploaded my data set in this particular Dropbox.", 'start': 246.847, 'duration': 4.625}, {'end': 255.956, 'text': "And once I execute this you will have a zip file and we'll use this part to unzip our files.", 'start': 251.632, 'duration': 4.324}], 'summary': 'Overview of using os, shutil, and glob modules for file handling, including data set upload and extraction from a dropbox location.', 'duration': 28.575, 'max_score': 227.381, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk227381.jpg'}, {'end': 662.303, 'src': 'embed', 'start': 639.043, 'weight': 3, 'content': [{'end': 646.149, 'text': "We're going to give a dense layer and, as we have number of classes as to so, our units over here will also be to activation.", 'start': 639.043, 'duration': 7.106}, {'end': 648.011, 'text': "that I'm going to use is going to be sigmoid.", 'start': 646.149, 'duration': 1.862}, {'end': 655.617, 'text': "And the input for this is going to be X this part is going to be the input for this and finally, let's create a model.", 'start': 649.492, 'duration': 6.125}, {'end': 662.303, 'text': "So we'll give the name here as models and we'll say model and just give the inputs and outputs.", 'start': 656.478, 'duration': 5.825}], 'summary': 'Creating a model with dense layer, 10 classes, and sigmoid activation.', 'duration': 23.26, 'max_score': 639.043, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk639043.jpg'}, {'end': 1013.685, 'src': 'embed', 'start': 983.321, 'weight': 4, 'content': [{'end': 985.322, 'text': 'So, as you can see, here we have 36.', 'start': 983.321, 'duration': 2.001}, {'end': 990.418, 'text': "if you're wondering why we have 36, that's the batch size that we have given here, and 26, 256,", 'start': 985.322, 'duration': 5.096}, {'end': 994.547, 'text': '256 is the size and 3 is the number of channel instead of 36..', 'start': 990.418, 'duration': 4.129}, {'end': 998.909, 'text': 'We can also give batch size as 64 this increase the number of images.', 'start': 994.547, 'duration': 4.362}, {'end': 1004.881, 'text': 'and now you can see our value here changes from 36 to 64.', 'start': 999.858, 'duration': 5.023}, {'end': 1008.442, 'text': "And now let's try to plot the images for this will create a function.", 'start': 1004.881, 'duration': 3.561}, {'end': 1013.685, 'text': "Let's say DF plot image and we are going to pass image array and then level.", 'start': 1008.503, 'duration': 5.182}], 'summary': 'Using batch size 36 and 64, the image array changes accordingly.', 'duration': 30.364, 'max_score': 983.321, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk983321.jpg'}], 'start': 37.135, 'title': 'Training deep learning models with tensorflow and data processing', 'summary': 'Discusses training deep learning models using tensorflow to recognize images of cats and dogs, and covers data processing steps for building a model with dense layer classification. it also includes visualizing image data with 25,000 images and setting dimensions and batch size for model training.', 'chapters': [{'end': 226.941, 'start': 37.135, 'title': 'Deep learning model with tensorflow', 'summary': 'Discusses the process of training a deep learning model using tensorflow to recognize images of cats and dogs, using a feature map to interpret the model, and utilizing tools like google collab, tensorflow, numpy, and matplotlib for implementation.', 'duration': 189.806, 'highlights': ['The process of training a model with 25,000 images of cats and dogs to create a feature map for recognizing patterns, and visualizing the gradients to interpret the deep learning model. The model is trained with 25,000 images of cats and dogs to create a feature map for pattern recognition, and the gradients are visualized to interpret the deep learning model.', "Utilizing pre-trained models like Inception or VGG 19 for the deep learning model, followed by testing the model's performance and using the grad cam method to predict a feature map in the form of a heat map. Pre-trained models like Inception or VGG 19 are used for the deep learning model, followed by testing the model's performance and using the grad cam method to predict a feature map in the form of a heat map.", 'Utilizing Google Collab for free GPU service and frameworks such as TensorFlow, numpy, and matplotlib for implementing the deep learning model. Google Collab is used for free GPU service, and frameworks like TensorFlow, numpy, and matplotlib are utilized for implementing the deep learning model.']}, {'end': 848.308, 'start': 227.381, 'title': 'Data processing and model building', 'summary': 'Covers data processing steps including file handling using os, file copying with shuttle, pattern matching with glob, and pre-processing images using inception v3 model, with the aim of building a model with a dense layer for classification of cat and dog images.', 'duration': 620.927, 'highlights': ['The chapter covers data processing steps including file handling using OS, file copying with Shuttle, pattern matching with Glob, and pre-processing images using Inception v3 model. The chapter includes information about handling files using OS, copying files using Shuttle, and pattern matching using Glob. It also covers pre-processing images using the Inception v3 model.', 'A model with a dense layer is built for the classification of cat and dog images. The chapter explains the process of building a model with a dense layer for the classification of cat and dog images. It provides details about using the Inception v3 model and setting up the dense layer for classification.']}, {'end': 1131.448, 'start': 848.828, 'title': 'Visualizing image data and model training', 'summary': 'Covers the process of visualizing image data with 25,000 images, setting dimensions and batch size for model training, and creating a function to plot and display image data.', 'duration': 282.62, 'highlights': ['The chapter covers the process of visualizing image data with 25,000 images, setting dimensions and batch size for model training, and creating a function to plot and display image data.', 'The batch size for model training is set at 64, increasing the number of images.', 'The image data consists of 25,000 images, with each class having 25,000 images, and the index values representing different categories.', 'A function is created to plot and display image data, allowing visualization of images of cats, dogs, and others, with the option to increase the number of images.']}], 'duration': 1094.313, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk37135.jpg', 'highlights': ["Utilizing pre-trained models like Inception or VGG 19 for the deep learning model, followed by testing the model's performance and using the grad cam method to predict a feature map in the form of a heat map.", 'The process of training a model with 25,000 images of cats and dogs to create a feature map for recognizing patterns, and visualizing the gradients to interpret the deep learning model.', 'The chapter covers data processing steps including file handling using OS, file copying with Shuttle, pattern matching with Glob, and pre-processing images using Inception v3 model.', 'A model with a dense layer is built for the classification of cat and dog images. The chapter explains the process of building a model with a dense layer for the classification of cat and dog images.', 'The chapter covers the process of visualizing image data with 25,000 images, setting dimensions and batch size for model training, and creating a function to plot and display image data.']}, {'end': 1557.016, 'segs': [{'end': 1159.653, 'src': 'embed', 'start': 1132.148, 'weight': 2, 'content': [{'end': 1135.729, 'text': "So now what we'll do can just turn off the axis that's done.", 'start': 1132.148, 'duration': 3.581}, {'end': 1137.57, 'text': 'Yep So this is all that we can do over here.', 'start': 1136.049, 'duration': 1.521}, {'end': 1141.997, 'text': "and let's now move ahead and let's try to create a model or train a model.", 'start': 1138.291, 'duration': 3.706}, {'end': 1142.978, 'text': "So let's do that.", 'start': 1142.277, 'duration': 0.701}, {'end': 1152.412, 'text': 'So that is from Keras not callbacks import model checkpoint and early stopping.', 'start': 1144.24, 'duration': 8.172}, {'end': 1155.411, 'text': "and now we'll just create a class instance of this.", 'start': 1153.07, 'duration': 2.341}, {'end': 1159.653, 'text': 'So for model checkpoint is going to be MC and we have model checkpoint.', 'start': 1155.871, 'duration': 3.782}], 'summary': 'Preparing to create and train a model using keras with model checkpoint and early stopping.', 'duration': 27.505, 'max_score': 1132.148, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1132148.jpg'}, {'end': 1349.843, 'src': 'embed', 'start': 1291.483, 'weight': 0, 'content': [{'end': 1297.545, 'text': "Make sure your runtime is in GPU or else it's definitely going to take quite a lot of time to train your CNN model.", 'start': 1291.483, 'duration': 6.062}, {'end': 1299.086, 'text': 'Meanwhile, that is training.', 'start': 1298.086, 'duration': 1}, {'end': 1301.347, 'text': 'Let me just give a proper alignment here.', 'start': 1299.306, 'duration': 2.041}, {'end': 1303.726, 'text': 'And yes, this should be fine.', 'start': 1302.687, 'duration': 1.039}, {'end': 1325.384, 'text': 'So as you can see here, we have done a small mistake.', 'start': 1322.802, 'duration': 2.582}, {'end': 1331.429, 'text': "So, although the model has trained perfectly fine, but the issue is that our model checkpoint didn't work,", 'start': 1325.724, 'duration': 5.705}, {'end': 1333.811, 'text': 'and the reason for this is because I had done a small typo.', 'start': 1331.429, 'duration': 2.382}, {'end': 1338.734, 'text': "So what we're going to do here is let me quickly take this up, validation, accuracy,", 'start': 1334.171, 'duration': 4.563}, {'end': 1343.478, 'text': 'and let me come back here and let me fix this up and let me do the same over here.', 'start': 1338.734, 'duration': 4.744}, {'end': 1347.741, 'text': "What happens over here is if I don't try to do this our model might end up overfitting.", 'start': 1343.958, 'duration': 3.783}, {'end': 1349.843, 'text': 'Let me do a cross check here once again.', 'start': 1348.102, 'duration': 1.741}], 'summary': 'Training cnn model on gpu to avoid overfitting issue.', 'duration': 58.36, 'max_score': 1291.483, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1291483.jpg'}, {'end': 1561.018, 'src': 'embed', 'start': 1536.952, 'weight': 3, 'content': [{'end': 1544.221, 'text': 'So, for every Epoch, as you can see, our accuracy has increased and for every loss, you know, for every Epoch, our losses also decreased.', 'start': 1536.952, 'duration': 7.269}, {'end': 1548.346, 'text': 'if you want the same thing to be done for our loss, to see how our Epoch has worked.', 'start': 1544.221, 'duration': 4.125}, {'end': 1551.109, 'text': 'So let me add this and execute this now.', 'start': 1548.746, 'duration': 2.363}, {'end': 1557.016, 'text': 'So as you can see right for every Epoch our loss value has decreased and this is the notes below in person now.', 'start': 1551.489, 'duration': 5.527}, {'end': 1561.018, 'text': 'So this is all there is about, you know, building a model and creating predictions,', 'start': 1557.376, 'duration': 3.642}], 'summary': 'Accuracy increased for every epoch, losses decreased for every epoch, showing model improvement.', 'duration': 24.066, 'max_score': 1536.952, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1536952.jpg'}], 'start': 1132.148, 'title': 'Implementing and training models with keras and gpu', 'summary': "Covers implementing model training with keras, setting up model checkpoint and early stopping callbacks, defining parameters, and executing model training with test data and callbacks. it also highlights training a cnn model on a gpu, emphasizing the importance of gpu for faster training, identifying and fixing mistakes in model training, retraining the model, and analyzing the model's performance through loss and accuracy visualization.", 'chapters': [{'end': 1290.482, 'start': 1132.148, 'title': 'Implementing model training with keras', 'summary': 'Covers the implementation of model training using keras, including setting up model checkpoint and early stopping callbacks, with a focus on monitoring validation accuracy, defining parameters, and executing the model training with test data and callbacks.', 'duration': 158.334, 'highlights': ['The chapter covers the implementation of model training using Keras, including setting up model checkpoint and early stopping callbacks.', 'The parameters for model checkpoint include file path for saving the best model, monitoring validation accuracy, and setting verbosity.', 'The parameters for early stopping include monitoring validation accuracy, defining minimum Delta, setting patience, and verbosity.', 'The process involves fixing callbacks in an array, training the model with test data and steps for Epoch, and specifying the number of Epochs with callbacks for model training.']}, {'end': 1557.016, 'start': 1291.483, 'title': 'Training cnn model on gpu', 'summary': "Highlights the process of training a cnn model on a gpu, including the importance of gpu for faster training, identifying and fixing mistakes in model training, retraining the model, and analyzing the model's performance through loss and accuracy visualization.", 'duration': 265.533, 'highlights': ['Training a CNN model on a GPU is essential for faster training and preventing overfitting, as it can significantly reduce the time required for training.', 'Identifying and fixing mistakes in model training, such as correcting typos and ensuring the proper inclusion of validation accuracy, is crucial for the successful training and functionality of the model.', 'Retraining the model requires rerunning the initialization process and then re-executing the training process, ensuring that the model starts training from the correct point and is saved properly.', "Analysis of the model's performance through visualization of loss and accuracy helps in understanding the training progress, where decreasing loss and increasing accuracy over epochs indicate the model's improvement and effectiveness."]}], 'duration': 424.868, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1132148.jpg', 'highlights': ['Training a CNN model on a GPU is essential for faster training and preventing overfitting, as it can significantly reduce the time required for training.', 'Identifying and fixing mistakes in model training, such as correcting typos and ensuring the proper inclusion of validation accuracy, is crucial for the successful training and functionality of the model.', 'The chapter covers the implementation of model training using Keras, including setting up model checkpoint and early stopping callbacks.', "Analysis of the model's performance through visualization of loss and accuracy helps in understanding the training progress, where decreasing loss and increasing accuracy over epochs indicate the model's improvement and effectiveness."]}, {'end': 2103.761, 'segs': [{'end': 1691.048, 'src': 'heatmap', 'start': 1557.376, 'weight': 0, 'content': [{'end': 1561.018, 'text': 'So this is all there is about, you know, building a model and creating predictions,', 'start': 1557.376, 'duration': 3.642}, {'end': 1568.063, 'text': "and if you're someone who is really curious about how well a model is working, what I can do is I can just create a function and pass an image,", 'start': 1561.018, 'duration': 7.045}, {'end': 1569.324, 'text': "but that's not our major agenda.", 'start': 1568.063, 'duration': 1.261}, {'end': 1575.828, 'text': "So our major agenda here first was to build a model and we'll train this on some let's say some class and over here.", 'start': 1569.724, 'duration': 6.104}, {'end': 1578.75, 'text': 'We are representing cat and dog and once I have this model,', 'start': 1575.848, 'duration': 2.902}, {'end': 1588.418, 'text': 'I will create a grad cam function which will take gradients from this deep learning model and it will create a heat map and whatever is my image is going to plot on top of that.', 'start': 1578.75, 'duration': 9.668}, {'end': 1591.02, 'text': "So let's not quickly move ahead and do that.", 'start': 1588.898, 'duration': 2.122}, {'end': 1598.867, 'text': "But just in case if you're someone who is really curious to see what's happening here, let me write a text validate our image.", 'start': 1592.421, 'duration': 6.446}, {'end': 1601.55, 'text': "So I've already written the code for this.", 'start': 1599.748, 'duration': 1.802}, {'end': 1602.871, 'text': 'Let me just copy that up.', 'start': 1601.95, 'duration': 0.921}, {'end': 1606.383, 'text': 'So as you can see I have written my function here.', 'start': 1603.982, 'duration': 2.401}, {'end': 1607.944, 'text': 'Let me give this in bold.', 'start': 1606.744, 'duration': 1.2}, {'end': 1613.728, 'text': "So now what we're going to do here is basically get a path load the image using load image function for this.", 'start': 1608.485, 'duration': 5.243}, {'end': 1617.15, 'text': 'We are supposed to import this from our libraries over here.', 'start': 1613.768, 'duration': 3.382}, {'end': 1620.712, 'text': "So we'll come back up pre-processing image or here.", 'start': 1617.53, 'duration': 3.182}, {'end': 1622.693, 'text': "We'll say load image and image story.", 'start': 1620.732, 'duration': 1.961}, {'end': 1624.756, 'text': "and we'll run this up.", 'start': 1623.795, 'duration': 0.961}, {'end': 1625.917, 'text': "So we'll scroll down.", 'start': 1625.097, 'duration': 0.82}, {'end': 1630.402, 'text': 'So we have loaded our image then we convert this into image to an array.', 'start': 1626.658, 'duration': 3.744}, {'end': 1634.467, 'text': 'So let me remove this and once is converted image to an array.', 'start': 1630.743, 'duration': 3.724}, {'end': 1639.307, 'text': 'So what we are going to do now is basically try to expand our model here.', 'start': 1635.385, 'duration': 3.922}, {'end': 1642.168, 'text': 'But before that we also are supposed to pre-process our image.', 'start': 1639.747, 'duration': 2.421}, {'end': 1645.93, 'text': 'So we have I is equal to pre-process input here.', 'start': 1642.388, 'duration': 3.542}, {'end': 1652.713, 'text': 'We are going to just pass the image that I wanted is I so pre-processing your input is pretty important and this part over here.', 'start': 1645.97, 'duration': 6.743}, {'end': 1654.434, 'text': "I'm just trying to increase the dimensions.", 'start': 1652.753, 'duration': 1.681}, {'end': 1656.535, 'text': 'So which we can see over here.', 'start': 1654.954, 'duration': 1.581}, {'end': 1660.417, 'text': "So let's now execute this says image is not defined.", 'start': 1657.195, 'duration': 3.222}, {'end': 1662.298, 'text': "Maybe that's because of this.", 'start': 1660.517, 'duration': 1.781}, {'end': 1671.005, 'text': "as you already loaded here, right? So we don't have to provide that and one more thing here is make sure your size is same.", 'start': 1663.138, 'duration': 7.867}, {'end': 1673.428, 'text': "It's going to be 256 comma 256.", 'start': 1671.146, 'duration': 2.282}, {'end': 1677.572, 'text': 'So as you can see here model has predicted it fine.', 'start': 1673.428, 'duration': 4.144}, {'end': 1684.378, 'text': "So images of a dog and the reason why you're getting this initial error is because we haven't regularized it.", 'start': 1678.192, 'duration': 6.186}, {'end': 1691.048, 'text': 'So if you want, you know, all you can do is divide this by 255 and So you cannot divide this by 255.', 'start': 1684.718, 'duration': 6.33}], 'summary': 'Building a model to predict cat and dog images. model predicted image of a dog with a 256x256 size.', 'duration': 56.352, 'max_score': 1557.376, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1557376.jpg'}, {'end': 1754.533, 'src': 'embed', 'start': 1724.318, 'weight': 2, 'content': [{'end': 1729.502, 'text': 'and, as you can see this, images of a cat, and a model has also predicted that this image is of a cat.', 'start': 1724.318, 'duration': 5.184}, {'end': 1731.243, 'text': 'and just one last thing before we move ahead', 'start': 1729.502, 'duration': 1.741}, {'end': 1733.905, 'text': 'So randomly, let me choose any value.', 'start': 1731.824, 'duration': 2.081}, {'end': 1738.169, 'text': 'Let me take copy path come up here and pass that particular image.', 'start': 1734.226, 'duration': 3.943}, {'end': 1740.707, 'text': "Let's now see what we get.", 'start': 1739.246, 'duration': 1.461}, {'end': 1744.328, 'text': 'So as you can see this is a dog and our model has predicted it.', 'start': 1741.187, 'duration': 3.141}, {'end': 1746.109, 'text': 'It says this model is of a dog.', 'start': 1744.388, 'duration': 1.721}, {'end': 1748.93, 'text': 'So now that we know what exactly is happening here.', 'start': 1746.529, 'duration': 2.401}, {'end': 1754.533, 'text': "So what we'll do is now we'll try to visualize what makes our deep learning model thing that this is of a dog.", 'start': 1749.41, 'duration': 5.123}], 'summary': 'Deep learning model predicted 2 images: 1 cat, 1 dog.', 'duration': 30.215, 'max_score': 1724.318, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1724318.jpg'}, {'end': 1792.023, 'src': 'embed', 'start': 1768.311, 'weight': 5, 'content': [{'end': 1774.794, 'text': "So if you're someone who's wondering what does this grad cam represents here? Well grad cam over here is basically class activation map.", 'start': 1768.311, 'duration': 6.483}, {'end': 1777.175, 'text': 'So gradient class activation map.', 'start': 1775.094, 'duration': 2.081}, {'end': 1779.056, 'text': "So that's what grad cam represents.", 'start': 1777.496, 'duration': 1.56}, {'end': 1782.298, 'text': 'Let me just put this in bold and execute this.', 'start': 1779.517, 'duration': 2.781}, {'end': 1783.759, 'text': 'So this is what grad cam is.', 'start': 1782.618, 'duration': 1.141}, {'end': 1785.4, 'text': "So what we're going to do?", 'start': 1784.179, 'duration': 1.221}, {'end': 1792.023, 'text': 'first off read the image and then we are going to extract the gradients from our image and then finally, we are going to display that on a function.', 'start': 1785.4, 'duration': 6.623}], 'summary': 'Grad cam is a class activation map used to extract gradients and display them on a function.', 'duration': 23.712, 'max_score': 1768.311, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1768311.jpg'}], 'start': 1557.376, 'title': 'Building deep learning model and image validation', 'summary': 'Explains building a deep learning model for image classification and discusses image validation techniques, including model predictions and grad cam visualization for evaluating performance.', 'chapters': [{'end': 1591.02, 'start': 1557.376, 'title': 'Building deep learning model', 'summary': "Explains the process of building a deep learning model for image classification, including creating predictions and generating a heat map using the grad cam function, to evaluate the model's performance.", 'duration': 33.644, 'highlights': ["The process involves building a deep learning model for image classification, with a focus on creating predictions and evaluating the model's performance.", "A function will be created to pass an image and generate a heat map using the grad cam function, providing a visual representation of the model's predictions.", 'The key agenda is to build a model for classifying images, specifically focusing on representing cats and dogs for training the model.']}, {'end': 2103.761, 'start': 1592.421, 'title': 'Image validation and grad cam visualization', 'summary': 'Discusses image validation using load image function and pre-processing, model predictions for dog and cat images, and the use of grad cam for class activation map visualization in deep learning.', 'duration': 511.34, 'highlights': ['The chapter discusses image validation using load image function and pre-processing.', 'Model predictions are made for dog and cat images, with the model correctly identifying the images.', 'The use of grad cam for class activation map visualization in deep learning is explained, including the process of extracting gradients from the image and displaying them.']}], 'duration': 546.385, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk1557376.jpg', 'highlights': ['The key agenda is to build a model for classifying images, specifically focusing on representing cats and dogs for training the model.', "The process involves building a deep learning model for image classification, with a focus on creating predictions and evaluating the model's performance.", 'Model predictions are made for dog and cat images, with the model correctly identifying the images.', 'The chapter discusses image validation using load image function and pre-processing.', "A function will be created to pass an image and generate a heat map using the grad cam function, providing a visual representation of the model's predictions.", 'The use of grad cam for class activation map visualization in deep learning is explained, including the process of extracting gradients from the image and displaying them.']}, {'end': 2629.803, 'segs': [{'end': 2172.896, 'src': 'embed', 'start': 2123.49, 'weight': 0, 'content': [{'end': 2126.172, 'text': "So finally what we're going to do is we're going to get the mean of this.", 'start': 2123.49, 'duration': 2.682}, {'end': 2128.013, 'text': "So we let's say pool.", 'start': 2126.712, 'duration': 1.301}, {'end': 2131.094, 'text': "Let's give the name here as pooled gradients or grads.", 'start': 2128.333, 'duration': 2.761}, {'end': 2141.04, 'text': "So we'll have TF dot reduce mean now, we'll just pass grads then we have axis is equal to 0 comma 1 comma 2.", 'start': 2131.575, 'duration': 9.465}, {'end': 2142.24, 'text': "That's because we have three channels.", 'start': 2141.04, 'duration': 1.2}, {'end': 2145.882, 'text': "And finally now we'll put this in the form of a heat map.", 'start': 2143.581, 'duration': 2.301}, {'end': 2148.264, 'text': "So we'll have lost convolution layer output.", 'start': 2146.183, 'duration': 2.081}, {'end': 2153.286, 'text': 'This would be the same thing but present at the index 0.', 'start': 2149.064, 'duration': 4.222}, {'end': 2154.567, 'text': 'and now we put this in a heat map.', 'start': 2153.286, 'duration': 1.281}, {'end': 2156.348, 'text': "Let's give the name here as heat map.", 'start': 2154.607, 'duration': 1.741}, {'end': 2168.174, 'text': 'So this will be last convolution layer output and then we will give this at symbol full grads and then finally we have TF dot new axis.', 'start': 2157.208, 'duration': 10.966}, {'end': 2172.896, 'text': "So now we'll do TF dot squeeze just to bring them in a shape.", 'start': 2169.835, 'duration': 3.061}], 'summary': 'Calculate mean of pooled gradients for 3 channels and create a heatmap.', 'duration': 49.406, 'max_score': 2123.49, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk2123490.jpg'}, {'end': 2287.545, 'src': 'embed', 'start': 2258.533, 'weight': 1, 'content': [{'end': 2261.254, 'text': "We're going to say save and display grad cam with the name of function.", 'start': 2258.533, 'duration': 2.721}, {'end': 2265.576, 'text': 'This takes an image path, will also have the heat map which you have created.', 'start': 2261.674, 'duration': 3.902}, {'end': 2269.298, 'text': 'then place where you want to save our grad cam image and then the alpha value.', 'start': 2265.576, 'duration': 3.722}, {'end': 2274.461, 'text': 'So they will have DF save and display grad cam.', 'start': 2269.679, 'duration': 4.782}, {'end': 2276.342, 'text': "So we'll have to take image path.", 'start': 2274.881, 'duration': 1.461}, {'end': 2279.839, 'text': 'and then we have heat map then we have camp path.', 'start': 2276.857, 'duration': 2.982}, {'end': 2287.545, 'text': "So you can just give any name will give it a scam dot JPG or you can give any name you want and then we'll also have the alpha value.", 'start': 2280.68, 'duration': 6.865}], 'summary': 'Function saves and displays grad-cam with image path, heat map, and alpha value.', 'duration': 29.012, 'max_score': 2258.533, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk2258533.jpg'}, {'end': 2405.827, 'src': 'embed', 'start': 2374.986, 'weight': 3, 'content': [{'end': 2378.289, 'text': 'So, as you can see, we have done this over here now and in the next stage.', 'start': 2374.986, 'duration': 3.303}, {'end': 2379.169, 'text': "what we're going to do?", 'start': 2378.289, 'duration': 0.88}, {'end': 2382.692, 'text': 'we are going to reimburse or we are going to superimpose heat map to our original image?', 'start': 2379.169, 'duration': 3.523}, {'end': 2385.914, 'text': 'So heat map basically has the gradients.', 'start': 2383.312, 'duration': 2.602}, {'end': 2391.058, 'text': 'This is what happens over here superimposed image that is heat map times alpha plus image.', 'start': 2386.255, 'duration': 4.803}, {'end': 2392.96, 'text': 'This is why we are using the alpha value.', 'start': 2391.319, 'duration': 1.641}, {'end': 2399.065, 'text': 'So this is what is happening here and finally coming down to you know, saving our model or saving a superimposed image.', 'start': 2393.36, 'duration': 5.705}, {'end': 2405.827, 'text': "This function here superimposed image dot, save and whatever the path you have given, that's a part where it's going to get saved.", 'start': 2399.685, 'duration': 6.142}], 'summary': 'Superimposing a heat map on an image using gradients and alpha value, then saving the superimposed image.', 'duration': 30.841, 'max_score': 2374.986, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk2374986.jpg'}], 'start': 2104.041, 'title': 'Visualizing cnn and grad cam function', 'summary': 'Discusses visualizing gradients in cnn using tensorflow, covering capturing gradients, calculating mean, creating a heat map, and generating grad cam images with specific alpha value and rgb color mapping, enhancing model interpretability and understanding.', 'chapters': [{'end': 2258.233, 'start': 2104.041, 'title': 'Visualizing convolutional neural network', 'summary': 'Explains the process of visualizing the gradients in a convolutional neural network using tensorflow, including capturing gradients, calculating mean, creating a heat map, and visualizing the result, with the mention of using tensorflow and official documentation for further queries.', 'duration': 154.192, 'highlights': ['Capturing gradients by using tape dot gradient and applying reduce mean to calculate the mean of the gradients, with axis as 0, 1, 2, representing the three channels.', 'Creating a heat map by manipulating the captured last convolution layer output and the gradients, with the use of squeeze and maximum functions in TensorFlow to form the heat map.', 'Mention of importing matplotlib and ipython display libraries for further visualization purposes.', 'Suggestion to refer to the official documentation of TensorFlow for a better understanding of the process and addressing further queries.']}, {'end': 2629.803, 'start': 2258.533, 'title': 'Function for grad cam image generation', 'summary': "Explains the creation of a function for generating grad cam images, demonstrating the process of superimposing heat maps on original images and visualizing the model's feature predictions, using a specific alpha value and rgb color mapping, resulting in improved model interpretability and understanding.", 'duration': 371.27, 'highlights': ["The function 'save and display grad cam' takes an image path, heat map, cam path, and alpha value, enabling the creation and display of Grad Cam images for model interpretability.", 'The process involves converting the input image to an array, rescaling the heat map, applying RGB color mapping, superimposing the heat map on the original image using the alpha value, and saving the superimposed image.', "The demonstration includes visualizing the heat map predictions on images of a dog and a human, showcasing the model's feature considerations and predictions, enhancing interpretability and understanding of the model's decisions."]}], 'duration': 525.762, 'thumbnail': 'https://coursnap.oss-ap-southeast-1.aliyuncs.com/video-capture/ejkRh9obVjk/pics/ejkRh9obVjk2104041.jpg', 'highlights': ['Creating a heat map by manipulating the captured last convolution layer output and the gradients, with the use of squeeze and maximum functions in TensorFlow to form the heat map.', "The function 'save and display grad cam' takes an image path, heat map, cam path, and alpha value, enabling the creation and display of Grad Cam images for model interpretability.", 'Capturing gradients by using tape dot gradient and applying reduce mean to calculate the mean of the gradients, with axis as 0, 1, 2, representing the three channels.', 'The process involves converting the input image to an array, rescaling the heat map, applying RGB color mapping, superimposing the heat map on the original image using the alpha value, and saving the superimposed image.']}], 'highlights': ['Training a CNN model on a GPU is essential for faster training and preventing overfitting, as it can significantly reduce the time required for training.', 'The process of training a model with 25,000 images of cats and dogs to create a feature map for recognizing patterns, and visualizing the gradients to interpret the deep learning model.', 'The chapter covers data processing steps including file handling using OS, file copying with Shuttle, pattern matching with Glob, and pre-processing images using Inception v3 model.', 'The session covers visualizing gradients using feature maps, starting with understanding the problem statement and workflow, discussing tools and frameworks, and finally jumping into the project.', "The process involves building a deep learning model for image classification, with a focus on creating predictions and evaluating the model's performance.", 'The chapter discusses image validation using load image function and pre-processing.', 'The key agenda is to build a model for classifying images, specifically focusing on representing cats and dogs for training the model.', 'The chapter covers the process of visualizing image data with 25,000 images, setting dimensions and batch size for model training, and creating a function to plot and display image data.', 'The use of grad cam for class activation map visualization in deep learning is explained, including the process of extracting gradients from the image and displaying them.', "The function 'save and display grad cam' takes an image path, heat map, cam path, and alpha value, enabling the creation and display of Grad Cam images for model interpretability."]}