##Pedestrian detection based on CenterNet
In this repo, we re-train the centernet on CityPerson dataset to get a pedestrian detector CenterNet
##Preparation
Please first install Anaconda and create an Anaconda environment using the provided package list.
conda create --name CenterNet --file conda_packagelist.txt
After you create the environment, activate it.
source activate CenterNet
Compiling Corner Pooling Layers
cd <CenterNet dir>/models/py_utils/_cpools/
python setup.py install --user
Compiling NMS
cd <CenterNet dir>/external
make
CityPerson dataset
- Download the CityPerson dataset and label files in images, label
- create a softlink in
datato your CityPerson dataln -s #to/yourdata/CityPerson data/
Training and Evaluation
To train CenterNet-52
python train.py --cfg_file CenterNet-52
The default configure in config/CenterNet-52.json is 2 (12G) GPUs and batchsize=12, you can modify them according to your case.
To evaluate your detector
python test.py --cfg_file CenterNet-52 --testiter #checkpoint_epoch
Demo
The demo images are stored in data/demo
python demo.py