Long Range Diffusion for Weakly Camouflaged Object Segmentation

This project provides the code and results for 'Long Range Diffusion for Weakly Camouflaged Object Segmentation'

Authors: Rui Wang, Caijuan Shi, Weixiang Gao, Changyu Duan, Ao Cai, Fei Yu, Yunchao Wei

Network Architecture

Preparation

The training and testing experiments are conducted using PyTorch with a single GeForce RTX 1080Ti GPU of 12 GB Memory.

Configuring your environment:

• Creating a virtual environment : conda create -n LRDNet python=3.9
• Installing necessary packages: pip install -r requirements.txt

Downloading Training and Testing Sets

• Download train datasets S-COD(COD10K-train+CAMO-train): TrainDatasets. "1" stands for foregrounds, "2" for backgrounds, and "0" for unlabeled regions.
• Download test datasets (CAMO-test+COD10K-test-test+NC4K ):TestDatasets

Pretrained Backbone Model

• Download pretrained backbone model:ResNet50, and put it in ./pth

Training

• Modify the dataset path in config.py.
• First Training: run python train.py, and it generates catalogue experiments\ with logs and weights.
• Second Training: run python train.py --ckpt=last --second_time
• You can also change the other config option by modify the config.py.

Testing Configuration

• Testing: run python test.py, and the result maps are in experiments\save_images\.
• We provide LRDNet testing maps and training weights presented in the papers.

Evaluation

• Tools: PySODMetrics A simple and efficient implementation of SOD metrics.

Results

Credit

The code is partly based on CRNet.