Retrieval-Robust-to-Object-Motion-Blur

Description

Pytorch code for paper: Retrieval Robust to Object Motion Blur

Accepted by ECCV 2024

Rong Zou, Marc Pollefeys and Denys Rozumnyi

Installation

The code is tested with Python 3.8.16.

Install this repository using the following commands:

# Clone the repository
git clone https://github.com/Rong-Zou/Retrieval-Robust-to-Object-Motion-Blur.git

# Change to the project directory
cd Retrieval-Robust-to-Object-Motion-Blur

# Install dependencies
pip install -r requirements.txt

Data Preparation

Download dataset zips from this link, and extract the data.

You may use the following commands:

# Download the dataset
wget https://cvg-data.inf.ethz.ch/romb/real_data.zip
wget https://cvg-data.inf.ethz.ch/romb/synthetic_data.zip
wget https://cvg-data.inf.ethz.ch/romb/synthetic_data_distractors.zip

# Unzip the dataset to ./data/, change the target path to your desired directory
unzip real_data.zip -d data/
unzip synthetic_data.zip -d data/
unzip synthetic_data_distractors.zip -d data/

See the dataset page for more details.

Pretrained Model

Download our pre-trained model from this link.

For testing, put the model in the directory same as the link file.

Testing

First modify the parameter values in the set_data_dirs.py script to configure the correct directories.

Test the model by running the testing script test.py:

python3 test.py

License

This project is licensed under the MIT License.

Citation

@inproceedings{blur_retrieval,
  author = {Rong Zou and Marc Pollefeys and Denys Rozumnyi},
  title = {Retrieval Robust to Object Motion Blur},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year = {2024}
}