EfficientNet-Light-head-RCNN for Human Detection

Google has recently released new efficient model architectures for edge devices, I thought it would be cool to see what the results of these models are as a backbone for FasterRCNNN. The EfficientNet-Light-head-RCNN has 31MAP on pedestrian detection and run at 12 FPS on my GTX 1060. Only EfficientNet B0 was tested.

Downloading the data

the crowdHuman Dataset can be downloaed here CrowdHuman

Data folder Structure (please create a data folder and make it like this)

==>data
===>annotations
===>Images
===>Images_test
====>Images_validation

COCO Evaluation Results

Average Precision (AP)	IoU=0.50:0.95	area= all	maxDets=100	0.310
Average Precision (AP)	IoU=0.50	area= all	maxDets=100	0.589
Average Precision (AP)	IoU=0.75	area= all	maxDets=100	0.295

demo

Pretrained Model

if you want to use the pretrained model, please download, create a folder named checkpoint and put it inside Pretrained_model

Training (once you have downlaoded the dataset)

python3 Train.py

Citing

@article{li2017light,
  title={Light-Head R-CNN: In Defense of Two-Stage Object Detector},
  author={Li, Zeming and Peng, Chao and Yu, Gang and Zhang, Xiangyu and Deng, Yangdong and Sun, Jian},
  journal={arXiv preprint arXiv:1711.07264},
  year={2017}
}

@article{shao2018crowdhuman,
    title={CrowdHuman: A Benchmark for Detecting Human in a Crowd},
    author={Shao, Shuai and Zhao, Zijian and Li, Boxun and Xiao, Tete and Yu, Gang and Zhang, Xiangyu and Sun, Jian},
    journal={arXiv preprint arXiv:1805.00123},
    year={2018}
}