Depth-Aware Concealed Crop Detection in Dense Agricultural Scenes (CVPR2024)

Overview

• Our ACOD-12K

• Our RISNet

Usage

The training and testing experiments are conducted using PyTorch with a single RTX 3090 GPU of 24 GB Memory.

Dependencies

• Creating a virtual environment in terminal: conda create -n RISNet python=3.8
• Installing necessary packages: pip install -r requirements.txt

Datasets

Our ACOD-12K can be found in Huggingface or Baidu Drive(0vy7)

Training

• The pretrained model is stored in Google Drive and Baidu Drive(51sr). After downloading, please change the file path in the corresponding code.

• You can use our default configuration to train your own RISNet, like this:

   python Train.py --epoch 100 --lr 1e-4 --batchsize 4 --trainsize 704 --train_path Your_dataset_path --save_path Your_save_path

Testing

• Our well-trained model is stored in Google Drive and Baidu Drive(4sgg). After downloading, please change the file path in the corresponding code.
• You can use our default configuration to generate the final prediction map, like this:

   python Test.py --testsize 704 --pth_path Your_checkpoint_path --test_path Your_dataset_path

Evaluation

• Matlab version: https://github.com/DengPingFan/CODToolbox
• Python version: https://github.com/lartpang/PySODMetrics

Results

Concealed Crop Detection(CCD)

• Results of our RISNet can be found in Google Drive and Baidu Drive(te1v).

Concealed Object Detection(COD)

• Results of our RISNet can be found in Google Drive and Baidu Drive(85oa).
• Our well-trained model can be found in Google Drive and Baidu Drive(9jw8).