💊 Mamba Capsule Routing Towards Part-Whole Relational Camouflaged Object Detection (MCRNet)

We introduce the Mamba to generate type-level mamba capsules from the pixel-level capsules for routing, which ensures a lightweight computation, further exploring the part-whole hierarchical relationships in COD.

---

📌 Environmental Setups

To set up your environment and install dependencies, run the following commands:

# Create virtual environment
conda create -n mcr python=3.10
conda activate mcr
conda install pytorch==2.1.0 torchvision==0.16.0 torchaudio==2.1.0 pytorch-cuda=11.8 -c pytorch -c nvidia
cd MCRNet
pip install -r requirements.txt

📌 Data Preparation

Download the camouflaged object detection datasets from Baidu, you can put datasets into the folder ./data/'. PIN: ss04

📌 Checkpoints

We offer the training weights of our MCRNet model on Baidu, which should be put into ./checkpoint/'. PIN: cs28

📌 Results

The prediction maps of our MCRNet can be found on Baidu. PIN: l27b