abd_net

ABD-Net: Attentive but Diverse Person Re-Identification

Input (query)

(Image from http://188.138.127.15:81/Datasets/Market-1501-v15.09.15.zip)

Shape : (batch, 3, height, width)

Output (Top 10 images from gallery images that are similar to query)

Usage

Automatically downloads the onnx and prototxt files on the first run. It is necessary to be connected to the Internet while downloading.

For the sample image,

$ python3 abd_net.py

If you want to specify the input image, put the image path after the --input option.
You can use --savepath option to change the name of the output file to save.

$ python3 abd_net.py --input IMAGE_PATH --savepath SAVE_IMAGE_PATH

If you want to specify the directory of gallery image, put the directory path after the --gallery_dir option.

$ python3 abd_net.py --gallery_dir gallery

Now, files in this gallery directory are very restricted.
Many more files can be found in the bounding_box_test directory of Market-1501-v15.09.15.zip or DukeMTMC-VideoReID.zip.

Once the program run, a intermediate result file containing the features of the gallery image will be created.
By adding the intermediate result file name after the --data option, you can use the intermediate result of the previous inference.

$ python3 abd_net.py --data result_resnet50.npy

By adding the model name after the --model option, you can specify the model.
The model name is selected from 'market1501', 'duke', 'msmt17'.

$ python3 abd_net.py --model market1501

Reference

• Attentive but Diverse Person Re-Identification

Framework

Pytorch

Model Format

ONNX opset=11

Netron

abd_net_market1501.onnx.prototxt
abd_net_duke.prototxt
abd_net_msmt17.prototxt