🚀 Course 🚀
Free: https://adataodyssey.com/xai-for-cv/
Paid: https://adataodyssey.com/courses/xai-for-cv/
In this video, we’re implementing Guided Backpropagation from scratch using PyTorch hooks—no external packages required! 💡 This hands-on approach gives you full flexibility and a deeper understanding of how CNNs make decisions.
We’ll explore three powerful ways to compute and interpret GBP gradients:
🎯 Target logit w.r.t. the input – pinpoints which input pixels drive a specific class prediction.
🧠 Target logit w.r.t. feature maps – shows how intermediate layers contribute to the final output.
🔬 Feature map element w.r.t. the input – reveals the spatial properties of high-level learned features.
By the end, you’ll not only understand how to implement GBP, but also how it compares to other model interpretation tools like Grad-CAM—and why it’s so useful for visualizing deep neural networks.
📌 Ideal for those working in:
Explainable AI (XAI)
Computer vision
Model interpretability
CNN visualization
Deep learning research
🚀 Useful playlists 🚀
XAI for CV: https://www.youtube.com/playlist?list=PLqDyyww9y-1QA4-o4tTAF_iD5cKCC1qEA
XAI: https://www.youtube.com/playlist?list=PLqDyyww9y-1SwNZ-6CmvfXDAOdLS7yUQ4
SHAP: https://www.youtube.com/playlist?list=PLqDyyww9y-1SJgMw92x90qPYpHgahDLIK
Algorithm fairness: https://www.youtube.com/playlist?list=PLqDyyww9y-1Q0zWbng6vUOG1p3oReE2xS
🚀 Get in touch 🚀
Medium: https://conorosullyds.medium.com/
Bluesky: https://bsky.app/profile/conorosullyds.bsky.social
Threads: https://www.threads.net/@conorosullyds
Website: https://adataodyssey.com/
🚀 Chapters 🚀
00:00 Introduction
01:46 Imports and model
04:45 Standard backpropagation
10:48 PyTorch Hooks
16:30 Target w.r.t. output
19:11 Intermediate layers
23:38 Elements in a layer
