TSCM: A Teacher-Student Model for Vision Place Recognition Using Cross-Metric Knowledge Distillation

Our work has been accepted by ICRA2024 👏 🎉

If you use our code in your work, please star our repo and cite our paper [pdf].

@inproceedings{shen2024icra,
	title={{TSCM: A Teacher-Student Model for Vision Place Recognition Using Cross-Metric Knowledge Distillation}},
	author={Shen, Yehui and Liu, Mingmin and Lu, Huimin and Chen, Xieyuanli},
	booktitle={IEEE International Conference on Robotics and Automation (ICRA)},
	year={2024}
}

Pittsburgh Dataset

You can download the pittsburgh dataset on https://www.dropbox.com/s/ynep8wzii1z0r6h/pittsburgh.zip?dl=0

How to use

If you want to verify the effect, you can download the stu_30k.pickle file and put it in the folder /logs/contrast/ ,run the following code

python vis.py

Train the pretrained models

In training mode

self.parser.add_argument('---split', type=str, default='val', help='Split to use', choices=['val', 'test'])

# train the teacher net
python main.py --phase=train_tea

# train the student net supervised by the pretrained teacher net
python main.py --phase=train_stu --resume=[logs/teacher_net_xxx/ckpt_best.pth.tar]

Evaluate the pretrained models

In test mode

self.parser.add_argument('---split', type=str, default='val', help='Split to use', choices=['val', 'test'])

needs to be changed to

self.parser.add_argument('---split', type=str, default='test', help='Split to use', choices=['val', 'test'])

You can run pre-trained models. The teacher_triplet/ckpt.pth.tar in the code needs to be changed to the appropriate name.

If this pre-training model doesn't work or you need more pre-training models, please contact me in Issue.

python main.py --phase=test_stu	 --resume=logs/teacher_triplet/ckpt.pth.tar

Use the pretrained model to PR

if you want to use the model to place recognition, you can replace the code in find_pair.py with the code in trainer.py and run the following code.

python main.py --phase=test_stu	 --resume=logs/teacher_triplet/ckpt.pth.tar

Thanks to the open source work of baseline, the code of TSCM is based on it.