Pytorch models definition and test for image retrieval

This is the implementation of the paper :

Portaz, M., Kohl, M., Chevallet, J. P., Quénot, G., & Mulhem, P. (2019). Object instance identification with fully convolutional networks. Multimedia Tools and Applications, 78(3), 2747-2764.

If you use it, please cite :

@article{portaz2019object,
	title={Object instance identification with fully convolutional networks},
	author={Portaz, Maxime and Kohl, Matthias and Chevallet, Jean-Pierre and Qu{\'e}not, Georges and Mulhem, Philippe},
	journal={Multimedia Tools and Applications},
	volume={78},
	number={3},
	pages={2747--2764},
	year={2019},
	publisher={Springer}
}

Test several approaches for images retrieval:

* Feature Extraction from Pretrained CNN
* Pretrained CNN finetuning
* Siamese network from scratch
* Siamese network with pretrained network

TrainClassif

Finetune a CNN for classification over few examples
Finetune only the classifier or the entire network

TrainSiamese

Train a siamese network with pairs selection for image retrieval