To assign a single emotion label to the video clip from the eight universal emotions (Calm, Happy, Sad, Angry, Fearful, Disgust, Surprised and Neutral) on the given multimodal dataset. Use the first 20 actors for the training and rest(04) for testing.
Approach: ConvNet (Spatial features) + LSTM (temporal features)
Limitations of Baseline:
- Capitalizes only on Facial Features rather than the combination of Spatial as well as Temporal
References -
- Paper ref. - https://sci-hub.se/10.1145/2818346.2829994
- The Ryerson Audio-Visual Database of Emotional Speech and Song(RAVDESS) Dataset - https://zenodo.org/record/1188976