Young Se Kim, Melnita Dabre, Chinmay Shirore, Seung Suk Lee
This project looks at a new approach for sentimental visual captioning and a new dataset for affective visual question answering.
Generic_SentiQA dataset can be found in the data folder. Please run download.sh to download all the required images for using this dataset.
Baseline includes jupiter notebooks for implementation of LSTM and OmniNet for sentimental visual captioning.
Approach includes jupiter notebooks for implementation of our new approaches for visual captioning including LSTM+LXMERT, gpt2 and file for preprocessing images.
Evaluation includes jupiter notebooks to evaluate and return evaluation metrics of BLUE, ROUGE-L, METEOR, CIDEr and SPICE with good and bad result of generated sentences with corresponding images.