Interpreting model YoloV5 by Grad-cam

[hoangkhoiLE]

🚀 Feature

I want to know more about the feature of model yolov5 by a technique Grad-cam
REF (https://keras.io/examples/vision/grad_cam/) or another method

Motivation

Understand more about feature, and to combine the result heatmap with another model (detection motion, colors) and to draw a heatmap about the zone of object detected

Thank you for your excellent work !

[github-actions]

👋 Hello @hoangkhoiLE, thank you for your interest in 🚀 YOLOv5! Please visit our ⭐️ Tutorials to get started, where you can find quickstart guides for simple tasks like Custom Data Training all the way to advanced concepts like Hyperparameter Evolution.

If this is a 🐛 Bug Report, please provide screenshots and minimum viable code to reproduce your issue, otherwise we can not help you.

If this is a custom training ❓ Question, please provide as much information as possible, including dataset images, training logs, screenshots, and a public link to online W&B logging if available.

For business inquiries or professional support requests please visit https://www.ultralytics.com or email Glenn Jocher at [email protected].

Requirements

Python 3.8 or later with all requirements.txt dependencies installed, including torch>=1.7. To install run:

$ pip install -r requirements.txt

Environments

YOLOv5 may be run in any of the following up-to-date verified environments (with all dependencies including CUDA/CUDNN, Python and PyTorch preinstalled):

• Google Colab Notebook with free GPU:
• Kaggle Notebook with free GPU: https://www.kaggle.com/ultralytics/yolov5
• Google Cloud Deep Learning VM. See GCP Quickstart Guide
• Docker Image https://hub.docker.com/r/ultralytics/yolov5. See Docker Quickstart Guide

Status

If this badge is green, all YOLOv5 GitHub Actions Continuous Integration (CI) tests are currently passing. CI tests verify correct operation of YOLOv5 training (train.py), testing (test.py), inference (detect.py) and export (export.py) on MacOS, Windows, and Ubuntu every 24 hours and on every commit.

[glenn-jocher]

@hoangkhoiLE I think gradcam may be more appropriate for classification models, but I could be wrong. I can't really help you here as no experience.

See Captum also for a model interpretability attempt by the pytorch authors (have not used myself either).
https://captum.ai/

[AndreaBrg]

@hoangkhoiLE Hi, did you ever succeeded in implementing Grad-Cam or Captum for YoloV5?

@glenn-jocher I see that this issue is linked under the Implemented Enhancements in the release 5.0 but I couldn't see anything related to the topic in the repo. Did I miss something?

[glenn-jocher]

@AndreaBrg ah sorry, the release changelog was built using an automated tool, I think it tries to categorize closed issues based on their tags, so this 'enhancement' tag probably led the issue to be grouped into the Implemented Enhancements section of the changelog.

In reality no theres been no update on this. I think detection vs classification with grad-cam would be quite a challenge due the complicated output of the detection models in relation to the classification models shown in gradcam. It's doable but someone would have to dedicate time and effort to investigate and develop this.

[AndreaBrg]

@glenn-jocher Thanks for the clarification.