Awesome
Frequency Perception Network for Camouflaged Object Detection
Authors: Runmin Cong, Mengyao Sun, Sanyi Zhang, Xiaofei Zhou, Wei Zhang, and Yao Zhao.
The source code can be found from Baidu Drive,CODE: MVPL
1. Preface
- This repository provides code for "Frequency Perception Network for Camouflaged Object Detection" ACM MM 2023. Paper
2. Proposed Method
2.1. Training/Testing
The training and testing experiments are conducted using PyTorch with one NVIDIA 2080Ti GPU of 32 GB Memory.
-
Configuring your environment (Prerequisites):
-
Installing necessary packages:
python 3.6
torch 1.11.0
numpy 1.22.4
mmcv-full 1.7.1
timm 0.6.13
mmdet 2.19.1
-
-
Downloading necessary data:
-
downloading dataset and move it into
./data/
, which can be found from Baidu Drive. -
downloading our weights and move it into
./snapshot/FPNet-GroupInsert/FPNet.pth
. -
downloading PVTv1-Large weights and move it into
./lib/models/pvt_large.pth
.
-
-
Training Configuration:
- After you download training dataset, just run
MyTrain_Val.py
to train our model.
- After you download training dataset, just run
-
Testing Configuration:
-
After you download all the pre-trained model and testing dataset, just run
MyTesting.py
to generate the final prediction maps. -
You can also download prediction maps ('CHAMELEON', 'CAMO', 'COD10K') from Baidu Drive.
-
2.2 Evaluating your trained model:
One evaluation is written in Python code please follow this the instructions in ./evaluator.py
and just run it to generate the evaluation results in. We implement four metrics: MAE (Mean Absolute Error), weighted F-measure, mean E-measure, S-Measure.