Deep Learning with PyTorch: Build, Train and Deploy an Image Classifier | Step-by-Step Tutorial

In this workshop, Alexey Grigorev, creator of the Machine Learning ZoomCamp, walks through how to build an image classification model in PyTorch from scratch using a fashion dataset as a real-world example. This session dives deep into the fundamentals of PyTorch, transfer learning, and model optimization, contrasting the framework’s flexibility and low-level control with the simplicity of Keras. You’ll learn about: Setting up your PyTorch environment on Google Colab with GPU acceleration Loading and preprocessing images using Pillow, NumPy, and Torchvision Leveraging pretrained models (MobileNet V2) and applying transfer learning Writing a custom Dataset class and DataLoaders for batching and shuffling Building and training a 10-class image classifier from the MobileNet base Implementing checkpointing, dropout, and data augmentation to prevent overfitting Exporting the final model to ONNX format for serverless deployment Links: Course: https://github.com/DataTalksClub/mach... Workshop: https://github.com/alexeygrigorev/wor... Colab: https://colab.research.google.com/ TIMESTAMPS: 00:00 Intro to PyTorch, Image Classification, and ML ZoomCamp 04:10 PyTorch vs Keras: Complexity and Popularity Explained 07:45 Loading Images and Converting to NumPy and Tensors 10:50 Using MobileNet V2 and ImageNet Pretrained Models 14:05 Image Preprocessing: Resize, Crop, and Normalize 17:10 Unsqueeze Explained: Prepping Inputs for Batching 21:00 From ImageNet to Fashion: Transfer Learning Setup 24:45 Building a Custom PyTorch Dataset Class 27:50 Creating DataLoaders for Batching and Shuffling 31:00 Freezing MobileNet and Adding a 10-Class Output Layer 34:30 Writing the PyTorch Training Loop vs model.fit 39:45 Running Validation and Measuring Model Accuracy 44:15 Tuning the Learning Rate for Better Training 48:40 Model V2: Adding an Inner Layer for Optimization 52:40 Saving Progress with Model Checkpointing 01:00:30 Regularization V3: Using Dropout to Prevent Overfitting 01:05:50 Regularization V4: Data Augmentation for More Training Data 01:10:25 Why Validation Data Shouldn’t Use Augmentation 01:17:00 Loading Checkpoints and Making Final Predictions 01:21:55 Exporting to ONNX for Serverless Model Deployment This talk is ideal for data scientists, ML engineers, and AI enthusiasts eager to deepen their understanding of PyTorch and model deployment workflows. Whether you’re transitioning from Keras/TensorFlow or building production-ready models, this session provides practical, hands-on insights for mastering deep learning with PyTorch. Connect with DataTalks.Club: Join the community - https://datatalks.club/slack.html Subscribe to our Google calendar to have all our events in your calendar - https://calendar.google.com/calendar/... Check other upcoming events - https://lu.ma/dtc-events GitHub: https://github.com/DataTalksClub LinkedIn -   / datatalks-club   Twitter -   / datatalksclub   Website - https://datatalks.club/ Connect with Alexey Twitter -   / al_grigor   Linkedin -   / agrigorev   Check our free online courses: ML Engineering course - http://mlzoomcamp.com Data Engineering course - https://github.com/DataTalksClub/data... MLOps course - https://github.com/DataTalksClub/mlop... LLM course - https://github.com/DataTalksClub/llm-... Open-source LLM course: https://github.com/DataTalksClub/open... AI Dev Tools course: https://github.com/DataTalksClub/ai-d... 👉🏼 Read about all our courses in one place - https://datatalks.club/blog/guide-to-... 👋🏼 Support/inquiries If you want to support our community, use this link - https://github.com/sponsors/alexeygri... If you’re a company, reach us at [email protected] #PyTorch #MachineLearning #DeepLearning #ImageClassification #TransferLearning #MobileNetV2 #Torchvision #DataAugmentation #Dropout #Checkpointing #ONNX #AIEngineering #MLZoomCamp #GoogleColab #NeuralNetworks #ComputerVision #KerasVsPyTorch #ModelDeployment #DeepLearningTutorial #PyTorchTutorial