https://github.com/Oulu-IMEDS/AdaTriplet

https://github.com/Oulu-IMEDS/AdaTriplet

Last synced: about 2 months ago
JSON representation

• Host: GitHub
• URL: https://github.com/Oulu-IMEDS/AdaTriplet
• Owner: Oulu-IMEDS
• Created: 2022-03-08T09:22:15.000Z (over 2 years ago)
• Default Branch: main
• Last Pushed: 2022-03-08T09:27:12.000Z (over 2 years ago)
• Last Synced: 2024-04-17T00:18:51.835Z (5 months ago)
• Language: Python
• Size: 1.53 MB
• Stars: 13
• Watchers: 4
• Forks: 0
• Open Issues: 2
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# AdaTriplet

## Dependencies

We provide file `environment.yml` for dependencies

## Installation

      conda env create -f environment.yml

      conda activate AdaTriplet

      cd AdaTriplet/

      pip install -e .

## Experiments

### Dataset

1. Knee X-ray: Download the OAI dataset via https://nda.nih.gov/oai/.

2. Chest X-ray: Download the ChestXrays-14 dataset at https://nihcc.app.box.com/v/ChestXray-NIHCC/ 

3. Create the image folder `./ResizedImages_{OAI|CXR}`

### Training

Run the script as follows:

      python train.py data_type= \ 

            image_crop_path= \  

            image_raw_path= \

            metadatapath= \

            datapath= \

            method=

Our code supports 8 `method_name`s: 

1. `AdaTriplet-AM` : AdaTriplet loss with AutoMargin selection

2. `AdaTriplet`

3. `Triplet-AM`: Triplet loss with AutoMargin selection

4. `Triplet`

5. `SCT`

6. `WAT`

7. `ArcFace`

8. `SoftTriplet`

###Test

Run the script as follows:

      python test.py data_type= 

      query_time=

      pretrained_matching_model_folder_path=

## Results

![Results on OAI](images/suppl_table4.png "Results on OAI")

![Results on CXR](images/suppl_table5.png "Results on CXR")

## Forensic Matching Results

![Results on CXR](images/suppl_fig2.png "Results on CXR")