chledowski/dbtpnet

DBTPNet

Introduction

This is the official implementation of the paper "Deep Bidirectional and Triplet-based Pedestrian Re-identification Network" (DBTPNet) in PyTorch.

Requirements

To install requirements: pip install -r requirements.txt

Samples

We have included one sample, it is located under this drive link: https://drive.google.com/file/d/1k9KgOb7tydhwfbSkHyji1vPn9kHhv7dA/view?usp=drive_link

Passing the sample through the model

To pass the sample through the model, run this command: PYTHONPATH=. python src/main.py

Before running the command, make sure that:

• the sample is downloaded to the folder samples/
• the normalization constants are updated in config.yaml if you're using your own sample

Outputs

The outputs of the model are localised under outputs/ after passing the sample through the model.

Visualization

DBT slice 0:

DBT slice 14:

DBT slice 29:

C-View:

Synthetic 2D:

References

To cite this work, please use

J. Chledowski, J. Park, and K. Geras, 
"Exploring synthesizing 2D mammograms from 3D digital breast tomosynthesis images", 
DICTA2023.