• Speech synthesis to help the blind people
• Automatically generate image description on e-commerce platforms
• Video description. Used for searching video

The notebooks for preprocessing data, training, evaluating the experiments are in strategy-1 branch

Read our thesis here

Simulate a sport event where the crowd is cheering, and the commentator is delivering his speech based on the situation of the match.

Demo video

• Resnet152 + LSTM, without Attention
• Downloads:
  • encoder
  • decoder

conda env create -f environment.yml

cd cv-nlp-end-term
streamlit run test.py