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.
Link to performance table
name: "input" , shape: [1x3x320x544] - An input image in the format [BxCxHxW], where:
Expected color order is BGR.
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 classx_min
, y_min
) - coordinates of the top left bounding box cornerx_max
, y_max
) - coordinates of the bottom right bounding box corner.[*] Other names and brands may be claimed as the property of others.