Pedestrian detection network based on SSD framework with tuned MobileNet v1 as a feature extractor. Some layers of MobileNet v1 are binary and use I1 arithm
Metric | Value |
---|---|
Average Precision (AP) | 84% |
Target pedestrian size | 60 x 120 pixels on Full HD image |
Max objects to detect | 200 |
GFlops | 0.750 |
GI1ops | 2.086 |
MParams | 1.165 |
Source framework | PyTorch* |
Average Precision metric described in: Mark Everingham et al. The PASCAL Visual Object Classes (VOC) Challenge.
Tested on an internal dataset with 1001 pedestrian to detect.
Name: input
, shape: [1x3x384x672] - An input image in the format [BxCxHxW], where:
The net outputs blob with shape: [1, 1, N, 7], where N is the number of detected bounding boxes. Each detection has the format [image_id
, label
, conf
, x_min
, y_min
, x_max
, y_max
], where:
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.
The net is tuned from pedestrian-detection-adas-0002.