Grad-CAM: Gradient-weighted Class Activation Mapping

Grad-CAM uses the class-specific gradient information flowing into the final convolutional layer of a CNN to produce a coarse localization map of the important regions in the image. It is a novel technique for making CNN more 'transparent' by producing visual explanations i.e visualizations showing what evidence in the image supports a prediction. You can play with Grad-CAM demonstrations at the following links:

Arxiv Paper Link: https://arxiv.org/abs/1610.02391

Grad-CAM VQA Demo: http://gradcam.cloudcv.org/vqa

Grad-CAM Classification Demo: http://gradcam.cloudcv.org/classification

Grad-CAM Captioning Demo: http://gradcam.cloudcv.org/captioning

Installing / Getting started

We use RabbitMQ to queue the submitted jobs. Also, we use Redis as backend for realtime communication using websockets.

All the instructions for setting Grad-CAM from scratch can be found here

Note: For best results, its recommended to run the Grad-CAM demo on GPU enabled machines.

Interested in Contributing?

Cloud-CV always welcomes new contributors to learn the new cutting edge technologies. If you'd like to contribute, please fork the repository and use a feature branch. Pull requests are warmly welcome.

if you have more questions about the project, then you can talk to us on our Gitter Channel.

Acknowledgements

• VQA_LSTM_CNN
• HieCoAttenVQA
• NeuralTalk2
• PyTorch