Skip to content

Latest commit

 

History

History
70 lines (52 loc) · 4.45 KB

README.md

File metadata and controls

70 lines (52 loc) · 4.45 KB

FoveaBox: Beyond Anchor-based Object Detector

This repo is a official implementation of "FoveaBox: Beyond Anchor-based Object Detector" on COCO object detection based on open-mmlab's mmdetection. Many thanks to mmdetection for their simple and clean framework.

Introduction

FoveaBox is an accurate, flexible and completely anchor-free object detection system for object detection framework, as presented in our paper https://arxiv.org/abs/1904.03797: Different from previous anchor-based methods, FoveaBox directly learns the object existing possibility and the bounding box coordinates without anchor reference. This is achieved by: (a) predicting category-sensitive semantic maps for the object existing possibility, and (b) producing category-agnostic bounding box for each position that potentially contains an object.

FoveaBox detection process.

Installation

This FoveaBox implementation is based on mmdetection. Therefore the installation is the same as original mmdetection.

Please check INSTALL.md for installation instructions.

Train and inference

The FoveaBox config is in configs/foveabox.

Inference

# single-gpu testing
python tools/test.py ${CONFIG_FILE} ${CHECKPOINT_FILE} [--out ${RESULT_FILE}] --eval bbox [--show]

# multi-gpu testing
./tools/dist_test.sh ${CONFIG_FILE} ${CHECKPOINT_FILE} ${GPU_NUM} [--out ${RESULT_FILE}] --eval bbox

Training

# single-gpu training
python tools/train.py ${CONFIG_FILE}

# multi-gpu training
./tools/dist_train.sh ${CONFIG_FILE} ${GPU_NUM} [optional arguments]

Please check GETTING_STARTED.md for detailed instructions.

Main Results

Results on R50/101-FPN with backbone

Backbone Style align ms-train Lr schd Mem (GB) Train time (s/iter) Inf time (fps) box AP Download
R-50 pytorch N N 1x 5.7 0.450 13.5 36.5 model
R-50 pytorch N N 2x - - 36.9 model
R-50 pytorch Y N 2x - - 37.9 model
R-50 pytorch Y Y 2x - - 40.1 model
R-101 pytorch N N 1x 9.4 0.712 11.5 38.5 model
R-101 pytorch N N 2x - - - 38.5 model
R-101 pytorch Y N 2x - - - 39.4 model
R-101 pytorch Y Y 2x - - - 41.9 model

[1] 1x and 2x mean the model is trained for 12 and 24 epochs, respectively.
[2] Align means utilizing deformable convolution to align the cls branch.
[3] All results are obtained with a single model and without any test time data augmentation.
[4] We use 4 NVIDIA Tesla V100 GPUs for training.

Any pull requests or issues are welcome.

Citations

Please consider citing our paper in your publications if the project helps your research. BibTeX reference is as follows.

@article{kong2019foveabox,
  title={FoveaBox: Beyond Anchor-based Object Detector},
  author={Kong, Tao and Sun, Fuchun and Liu, Huaping and Jiang, Yuning and Shi, Jianbo},
  journal={arXiv preprint arXiv:1904.03797},
  year={2019}
}