metadata
datasets:
- zeahub/echonet-dynamic
language:
- en
base_model:
- MCG-NJU/videomae-base
tags:
- medical
This repository provides a VideoMAE-based Segmentation Model. It includes:
- Backbone: A ViT-Base encoder pre-trained with VideoMAE on EchoNet-Dynamic.
- Segmentation Head: A decoder designed to predict pixel-wise masks for the Left Ventricle at the end of systole and end of diastole. The input is a 16-frame video.
The coder is published via github: https://github.com/mar1lle/echo-segmentation