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
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.
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.