Anti-aliasing-Tiny-Object-Detection

This is the official repository for the paper 'The Importance of Anti-Aliasing in Tiny Object Detection'.

Requirements

python = 3.7.10
pytorch = 1.10.0
cuda = 10.2
numpy = 1.21.2
mmcv-full = 1.4.7 
mmdet = 2.19.0

Dataset

TinyPerson

Following SSPNet to prepare the TinyPerson.

WiderFace

Following vedadet to prepare the WiderFace.

DOTA

Following OBBDetection to prepare the DOTA.

Installation

Please refer to Installation for installation instructions.

Usage

Training

./tools/dist_train.sh configs/anti/bhwave_ch3.3.py 4 --cfg-options optimizer.lr=0.004 --work-dir work_dirs/bhwave_ch3.3/

Testing

./dist_test.sh configs/anti/bhwave_ch3.3.py work_dirs/bhwave_ch3.3/latest.pth 2 --format-only

More usage please refer to mmdetection.

Citation

If you use this codebase or idea, please cite our paper:

@ARTICLE{ning2023acml,
       author = {{Ning}, Jinlai and {Spratling}, Michael},
        title = {The Importance of Anti-Aliasing in Tiny Object Detection},
      journal = {arXiv e-prints},
     keywords = {Computer Science - Computer Vision and Pattern Recognition},
         year = {2023},
        month = {oct},
}

Acknowledgement

This work is developed based on the MMDetection, SSPNet, vedadet and OBBDetection.