Open-source schedule

• 28 Jun 2024: Open-sourced model and test code
• 19 Aug 2024: Released prediction map, pretrained model, backbone, and datasets
• Training source code will be released upon paper acceptance.

Downloads

Item	Baidu Pan	Google Drive
Prediction Map	Link	Link
Pretrained Model	Link	Link
Backbone Model	Link	Link
Datasets	Link	Link

environment

1. Python 3.8.13
2. others packages can be found at requirement.txt

start

1. git clone https://github.com/CHANTILLY2023/MAMIFNet

2. conda create --name MAMIFNet python=3.8.12

3. conda activate MAMIFNet

4. pip install torch==1.9.1+cu111 torchvision==0.10.1+cu111 torchaudio==0.9.1 -f https://download.pytorch.org/whl/torch_stable.html

5. pip install -r requirements.txt

6. pip install -U openmim

7. mim install mmcv-full==1.5.2

test

The relevant setting paths for data and weights are located in config.py. The default test is as follows:

1. create folder './data', download datasets file COD_dataset at https://pan.baidu.com/s/1Kpxaqv0n5YP-kDT8Id4wzA?pwd=pugv, unzip and move it to './data'

2. download backbone model at https://pan.baidu.com/s/16bKJaxewVfFXd73ICV6zDg?pwd=vn6v and unzip it.

3. download pretrained model at https://pan.baidu.com/s/10gWYyESaV6LaCZv8_1PVBA?pwd=z8d1 and unzip it.

4. run test.py