Crop and bound-box interesting parts of images using saliency maps generated with Attentive Convolutional LSTM Residual Network
