Gradient-weighted Class Activation Mapping [1] is a technique which produces visual explanations for decisions made by Convolutional Neural Networks without the need for architectural changes or retraining of the network under inspection.
This code offers a Grad-CAM based analysis for frozen graph models in TensorFlow 1. Images consisting of the heatmaps together with the original images are produced and stored in the defined image folder. More details about the implementation can be found in the comments of the code. The example images in the 'color' folder are taken from [2].
Clone the repository:
git clone https://github.com/lkreiskoether/grad-cam-analysis.gitpython grad-cam-analysis.py <path to model> <path to image folder> '[<x-dim images>,<y-dim images>]' <number of classes>python grad-cam-analysis.py model/frozen_graph.pb color '[700,700]' 5Example of one of the produced outputs running the example from above:
Thanks to https://github.com/jacobgil/keras-grad-cam and https://github.com/Ankush96/grad-cam.tensorflow for providing many building blocks for the code.
[1] R. R. Selvaraju et al., “Grad-CAM: Why did you say that? Visual Explanations from Deep Networks via Gradient-based Localization,” CoRR, vol. abs/1610.0, no. 2, pp. 336–359, 2016, doi: 10.1007/s11263-019-01228-7.
[2] P. Bergmann, M. Fauser, D. Sattlegger, and C. Steger, “MVTEC ad-A comprehensive real-world dataset for unsupervised anomaly detection,” in Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, 2019, vol. 2019-June, pp. 9584–9592, doi: 10.1109/CVPR.2019.00982.