Real-time kiosk user analysis system using pedestrian/face detection and age/gender classification
(Source : https://developer.nvidia.com/embedded/jetson-nano-developer-kit)
Pedestrian and face detection network using MobileNet and skip connection
Detection algorithm is constructed based on YOLOv2.
Age and gender classification network using Mobilenet and CBAM
Age is predicted in 9 super categories and 9 set of 5-way sub categories.
Evaluation using COCO 2014 validation dataset and AFAD-FULL test dataset
mAP |
---|
0.354 |
Age accuracy | Gender accuracy |
---|---|
84.13% | 97.32% |
Visualization using COCO 2014 validation dataset