Automatic Image Captioning
Employed a hybrid CNN-LSTM model with attention mechanism for the task of image captioning on the Flickr8k dataset.
Used transfer learning to obtain useful image vector embeddings and achieved a BLEU score of 0.18 on test set.