Introduction

A state of art detector for densely packed scenes dataset SKU-110K.

For more information, please read our technical report.

License

This project is released under the Apache 2.0 license.

Installation

1. To clone this project to your own device.

   git clone https://github.com/Media-Smart/SKU110K-DenseDet.git
   

2. Please refer to INSTALL.md for installation and dataset preparation.

3. Click here to download the weights.

Get Started

Please go to GETTING_STARTED.md for the basic usage of MMDetection.

Before you are about to run or train, please edit the path in configuration.

Test

python tools/test.py configs/SKU_fusion_bfp_x101_32x4d.py \
    ${YOUR_WEIGHT_PATH} \
    --show

Train

python tools/train.py configs/SKU_fusion_bfp_x101_32x4d.py

Performance

The performance applied on SKU-110k of our best solution.

	mAP	AP@50	AP@75
Val Set	58.0%	92.0%	66.5%
Test Set 1	58.7%	92.9%	67.3%

For more detail, please browse the configuration.

Contact

This repository is currently maintained by Tianze Rong (@SteelBeamR), Hongxiang Cai (@hxcai), Yichao Xiong (@mileistone).

Credits

The whole project is based on mmcv v0.5.9 and mmdetection v1.0rc1 via open-mmlab.