GradCAM

Reimplementation of https://arxiv.org/pdf/1610.02391.pdf in Tensorflow 2.1.

Authors

• Max Holmberg
• Joel Lidin

Data for sentiment classification

The datasets and word2vec files for the sentiment classification can be found here:

• https://www.kaggle.com/rtatman/glove-global-vectors-for-word-representation?select=glove.6B.200d.txt
• https://www.kaggle.com/c/movie-review-sentiment-analysis-kernels-only/data

Put the files in a folder called dataset-text with the following structure:

• dataset-text/train.tsv
• dataset-text/test.tsv
• dataset-text/glove.6B.200d.txt

Run Grad-CAM for Images

python run.py --path image_path --output output_path

Run Grad-CAM for binary sentiment classification

python run.py --text --output output_path --weights sentiment_cnn_weights/cp.ckpt --sentence "I love this movie"

Results

Grad-CAM for Image classification

Original	Heatmap overlay (predicted: Mountains)

Original	Heatmap overlay (predicted: Egyptian cat)

Original	Heatmap overlay (predicted: Curly coated retriever)

Grad-CAM for sentiment classification (predicted: positive)