Use Case and High-Level Description

This is a person detector for the ASL Recognition scenario. It is based on ShuffleNetV2-like backbone that includes depth-wise convolutions to reduce the amount of computation for the 3x3 convolution block and FCOS head.



Metric Value
Persons AP on COCO 80.0%
Minimal person height 100 pixel
GFlops 0.986
MParams 1.338
Source framework PyTorch*

Average Precision (AP) is defined as an area under the precision/recall curve.


Image, name: input, shape: 1, 3, 320, 320 in the format 1, C, H, W, where:

  • C - number of channels
  • H - image height
  • W - image width

Expected color order is BGR.


The net outputs blob with shape: 100, 5 in the format N, 5, where N is the number of detected bounding boxes. For each detection, the description has the format: [x_min, y_min, x_max, y_max, conf], where:

  • (x_min, y_min) - coordinates of the top left bounding box corner
  • (x_max, y_max) - coordinates of the bottom right bounding box corner
  • conf - confidence for the predicted class

Legal Information

[*] Other names and brands may be claimed as the property of others.