Implementation for Learnt Contrastive Concept Embeddings for Sign Recognition.
[07/04/2024] Created Repo for the paper titled "Learnt Contrastive Concept Embeddings for Sign Recognition", Work in progress with the aim of full release by end of May 2024.
[14/04/2024] Added snippet of the head module used for learnt concept embeddings. The contextual head can be added to existing sign recognition models along with the recognition loss (to be added).
Due to my current workload and deadlines, my aim is to add code bit by bit once a week. With goal of adding the code that contains main contributions of the paper first followed by the full training scripts.
[x] Add contextual head
[x] Add dropmask regularisation
[ ] Add full model
[ ] Add recognition loss
[ ] Add conceptual loss
[ ] Add keypoint approach
[ ] Add video approach
[ ] Add training script
[ ] Add evaluation