Created by Yuchong Li
This repository contains PyTorch implementation for MT-FiST.
We introduce a multi-task fine-grained spatio-temporal model (MT-FiST) to recognize surgical action triplets in CholecT50 Dataset.
Our code is based on MT-RCNet-CL and MC Loss.
The dataset and evaluation metrics are here.
We use the ResNet-50 as the backbone pre-trained on ImageNet-1K.
Our model weights and the .pkl file can be downloaded in link.