Code for the paper "Integrating Visual and Semantic Similarity Using Hierarchies for Image Retrieval"
The dataset is arranged such that each class has a directory with the corresponding images placed in them. An example directory structure is shown below.
├── dataset
│ ├── train_data
│ │ ├── class1
│ │ ├── class2
...
│ │ ├── classN
│ ├── test_data
│ │ ├── class1
│ │ ├── class2
...
│ │ ├── classN
Each dataset is followed by a csv file containing the class name and the corresponding classification label. An example for CIFAR10 is given in data/cifar10.csv.
The dataset paths and the id paths (csv files) should be included in the config.py.
The hyperparameters and arguments needed for training the network are available in config.py.
To launch the training, run
python3 train.py
The code automatically splits the dataset into train and validation.
To launch the inference, run
python3 main_img_retrieval.py
This computes the hierarchy and performs image retrieval.
If you use this code, please cite the following paper:
Aishwarya Venkataramanan, Martin Laviale, and Cédric Pradalier. "Integrating Visual and Semantic Similarity Using Hierarchies for Image Retrieval." International Conference on Computer Vision Systems. Cham: Springer Nature Switzerland, 2023.
@inproceedings{venkataramanan2023integrating,
title={Integrating Visual and Semantic Similarity Using Hierarchies for Image Retrieval},
author={Venkataramanan, Aishwarya and Laviale, Martin and Pradalier, C{\'e}dric},
booktitle={International Conference on Computer Vision Systems},
pages={422--431},
year={2023},
organization={Springer}
}