This repo is a re-implementation of the model published in One-Shot Texture Retrieval with Global Context Metric in ijcai2019.

Note that these code contain some of my own modifications / improvements, please do not confuse this repo with the original / official method!

To run model training on your own pc:

Navigate to the folder directory, open a terminal and create a virtual environment:

python3 -m venv env               # Create a virtual environment

source env/bin/activate           # Activate virtual environment

Install none pytorch dependencies:

pip install -r requirements.txt   # Install dependencies

Install pytorch 1.10 (you might have to use a different version depending on your CUDA version)

pip3 install torch==1.10.0+cu113 torchvision==0.11.1+cu113 torchaudio==0.10.0+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html

Install albumentations if you want to play with data augmentation:

pip install -U albumentations

To start the training:

python3 train_ver2.py

If you encounter some runtime memory issues, you can decrease the batch_size / num_workers according to your GPU spec

To test the trained model:

python3 localImageTest.py

Most likely you will be getting "file not found error" since I hardcored most of the file directories, remember to change them in the scripts according to your file system.

To exit virtual environment:

deactivate                       # Exit virtual environment

DTD dataset can be downloaded from here:

https://www.robots.ox.ac.uk/~vgg/data/dtd/

Please place the downloaded .pth file under the root directory (where train.py is placed) for the train.py to work.

The encoder backbone (pre-trained ResNet50 19c8e357) can be downloaded from here:

https://download.pytorch.org/models/resnet50-19c8e357.pth

Please place the downloaded .pth file under /utils/model/ for the train.py to work.