SKU110K DenseDet Save
A state of art detector for densely packed scenes dataset SKU-110K
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
-
To clone this project to your own device.
git clone https://github.com/Media-Smart/SKU110K-DenseDet.git -
Please refer to INSTALL.md for installation and dataset preparation.
-
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.
Open Source Agenda Badge
