Part 4: Convolutional Neural Networks – LeNet5 | Lesson: Pooling in Neural Networks
Join us for the "Practical Computer Vision with PyTorch and FiftyOne" workshop series. This is a 12-part, hands-on series that teaches you how to work with images, build and train models, and explore tasks like image classification, segmentation, object detection, and image generation. Each session combines straightforward explanations with practical coding in PyTorch and FiftyOne, allowing you to learn core skills in computer vision and apply them to real-world tasks.
In Part 4, In this session, we’ll explore CNN fundamentals by diving into the mechanics of convolutions and pooling. Participants will implement LeNet5 to grasp how basic convolutional layers operate, with practical insights on the features produced by convolutions using FiftyOne.
These are hands-on maker workshops that make use of GitHub Codespaces, Kaggle notebooks, and Google Colab environments, so no local installation is required (though you are welcome to work locally if preferred!)
Learn more about FiftyOne: https://github.com/voxel51/fiftyone
Check out the calendar for upcoming computer vision events: https://voxel51.com/computer-vision-e...
#computervision #ai #artificialintelligence #machinevision #machinelearning #datascience #opensource
Contents of this video --
00:00 - Introduction
00:10 - Purpose of Pooling: Reducing Size & Parameters
00:20 - Types of Pooling: Max, Mean, and Global
00:30 - Example Network Architecture Overview
00:48 - Expanding Feature Maps Through Convolutions
01:10 - Applying Max Pooling to Downsample Spatial Dimensions
01:25 - Transition from Feature Maps to Flattened Vector
01:37 - Classification via Dense Layers
01:52 - Max Pooling Explained: Preserving Strong Activations
02:20 - Mean Pooling Explained: Smoother Output with Averages
02:46 - Use Cases for Mean Pooling (Textures, Backgrounds)
02:59 - Global Pooling: Creating Fixed-Length Feature Vectors
03:22 - Adaptive Max and Mean Pooling
03:42 - Combining Global Pooling Outputs into a Vector
04:10 - Replacing Flattening with Global Pooling in Models
04:28 - Benefits: Fewer Parameters and Less Overfitting
04:45 - Use in Modern Architectures (e.g. ResNet)
05:00 - Pooling in Practice: Interpretability & Class Activation Maps
05:19 - Recap: Max, Mean, and Global Pooling Use Cases
05:55 - Pooling Impact on Model Size and Feature Retention
06:06 - Choosing the Right Pooling Method for Your Task
--