← Aman Kumar Limited · RegisterNotes to the accounts · EMOT
Archive · research · 2024 · Archived
Multimodal Emotion Recognition
Emotion classification across 8 classes with custom attention modules, plus an LSTM that predicts the next expression.
- Role
- Solo
- Solo
- Outcome
- Emotion classification across 8 classes with custom attention modules, plus an LSTM that predicts the next expression.
- Links
- Private codebase — described below
- Stack
- PyTorch · Vision Transformers · LSTM · LoRA · VGG-19
Custom attention modules (facial-region attention, micro-expression detection, cultural-context adaptation) classify 8 emotion classes. An LSTM models the sequence and predicts the next expression. Training used mixed precision and was monitored in TensorBoard.
This is a portfolio, not an offer of securities. No shares exist. Figures are about shipped work.