README.md

Project Introduction

This project is divided into two main parts, STAVOS_ Main is responsible for training the model, including the code for the semi supervised video object segmentation model, The STAVOS model can be used to segment general videos. STAVOS_ Medaka is responsible for the application of Medaka, which can segment the heartbeat video of Medaka, obtain its ventricular related parameters and visualization results.

Display of project results

folder introduce：

Data-medaka-Lateral-right folder: N datasets

Data-medaka-Ventural folder: R dataset

DAVIS-2016-trainval:

DAVIS-2017-trainval-480p:

input folder:the video to be inputted

output folder: The output result of the automation system

STAVOS_main:project code for training and testing the weights of the STAVOS model

STAVOS_medaka: automation system code for medaka ventricle

test_predict: Test results of the dataset

trained_model: Weights of trained models

data acquisition

Due to the upper limit on the file size uploaded by GitHub, the dataset and model weights cannot be uploaded here.

Data-medaka-Lateral-right folder: https://cowtransfer.com/s/d2546440518240

Data-medaka-Ventural folder: https://cowtransfer.com/s/3d0ed4b3d8ae4b

trained_model: https://cowtransfer.com/s/7c37a69ceedf4d

DAVIS-2016-trainval:https://davischallenge.org.

DAVIS-2017-trainval-480p:https://davischallenge.org.

How to use:

1, Download the required dataset and project code

2, configuration environment

Deep learning environment：conda install pytorch==1.12.0 torchvision==0.13.0 torchaudio==0.12.0 cudatoolkit=11.3 -c pytorch

Other libraries: Install whatever is missing.

3, Ensure the relative position of files

If not correct, follow the code prompts to make corrections

4, training and testing

--STAVOS_main

run "train_main.py" for model training

run "test_main.py" for model testing

You can also download the trained model from https://cowtransfer.com/s/6ec386cb595042, or contact email:2514819977@qq.com

5, application

--STAVOS_medaka

run "main.py" for medaka ventricular segmentation and parameter acquisition