# vehicle-reid-0001¶

## Use Case and High-Level Description¶

This is a vehicle reidentification model for a general scenario. It uses a whole car 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 developed for fast inference. A single reidentification head from the 1/16 scale feature map outputs an embedding vector of 512 floats.

## Specification¶

Metric

Value

VeRi-776* rank-1

96.31 %

VeRi-776* mAP

85.15 %

Camera location

All traffic cameras

Support of occluded vehicles

YES

Occlusion coverage

<50%

GFlops

2.643

MParams

2.183

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.

## Input¶

### Original Model¶

One image of the shape 1, 3, 208, 208 in the B, C, H, W format, where:

• B - batch size

• C - channel

• H - height

• W - width

Channel order is RGB.

### Converted Model¶

One image of the shape 1, 3, 208, 208 in the B, C, H, W format, where:

• B - batch size

• C - channel

• H - height

• W - width

Channel order is BGR.

## Output¶

The net outputs a vector descriptor, which can be compared with other descriptors using the cosine distance.

### Original Model¶

Blob of the shape 1, 512 in the B, C format, where:

• B - batch size

• C - predicted descriptor size

### Converted Model¶

Blob of the shape 1, 512 in the B, C format, where:

• B - batch size

• C - predicted descriptor size

You can download models and if necessary convert them into OpenVINO™ IR format using the Model Downloader and other automation tools as shown in the examples below.

omz_downloader --name <model_name>
omz_converter --name <model_name>