This is a face and person detector for Retail scenario. It is based on MobileNetV2-like backbone that includes depth-wise convolutions to reduce the amount of computation for the 3x3 convolution block. The two SSD heads from 1/8 (for face detection) and 1/16 (for person detection) scale feature maps has 9 clustered prior boxes.
| Metric | Value |
|---|---|
| AP for persons | 88.12% |
| AP (WIDER) for faces | 83.59% (>64px), 87.55% (>100px) |
| Pose coverage | Standing upright, parallel to image plane |
| Support of occluded persons | YES |
| Occlusion coverage | <50% |
| GFlops | 2.757 |
| MParams | 0.791 |
| Source framework | Caffe* |
Average Precision (AP) is defined as an area under the precision/recall curve.
-
name: "input" , shape: [1x3x320x544] - An input image in the format [BxCxHxW], where:
- B - batch size
- C - number of channels
- H - image height
- W - image width
Expected color order is BGR.
- The net outputs blob with shape: [1, 1, N, 7], where N is the number of detected
bounding boxes. For each detection, the description has the format:
[
image_id,label,conf,x_min,y_min,x_max,y_max]image_id- ID of the image in the batchlabel- predicted class IDconf- confidence for the predicted class- (
x_min,y_min) - coordinates of the top left bounding box corner - (
x_max,y_max) - coordinates of the bottom right bounding box corner.
[*] Other names and brands may be claimed as the property of others.