person-reidentification-retail-0265

Use Case and High-Level Description

This is a person reidentification model for a general scenario. It uses a whole body image as an input and outputs an embedding vector to match a pair of images by the cosine distance. The model is based on the OmniScaleNet backbone with Linear Context Transform (LCT) blocks developed for fast inference. A single reidentification head from the 1/16 scale feature map outputs an embedding vector of 256 floats.

Example

person-reidentification-retail-0265.jpg

Specification

Metric Value
Market-1501 rank@1 accuracy 95.9 %
Market-1501 mAP 87.7 %
Pose coverage Standing upright, parallel to image plane
Support of occluded pedestrians YES
Occlusion coverage <50%
GFlops 1.993
MParams 2.103
Source framework PyTorch*

The cumulative matching curve (CMC) at rank-1 is accuracy denoting the possibility to locate at least one true positive in the top-1 rank. Mean Average Precision (mAP) is the mean across Average Precision (AP) of all queries. AP is defined as the area under the precision and recall curve.

Performance

Inputs

The net expects one input image of the shape [1x3x256x128] in the [BxCxHxW] format, where:

The expected color order is BGR.

Outputs

The net outputs a blob with the [1, 256] shape named descriptor, which can be compared with other descriptors using the cosine distance.

Legal Information

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