Visual Attention | Image Captioning | Visualization | tensorflow | python

Image Caption Generation with Visual Attention using tensorflow / python
- Using InceptionV3 to as CNN-Encoder to extract features from images
- Using BahdanauAttention as part of RNN-Decoder
- Trained on MS-COCO 2017 dataset with 20 epochs only
- Predicted captions are shown along with attention-weights overlayed on the test input images
- Parts of an image that get higher attention for a given word (in the caption predicted) are brighter

(#python #imageprocessing #mscoco #attention #inception #captiongeneration #rnn #cnn #cv #vision )