Skip to content

Latest commit

 

History

History
78 lines (44 loc) · 1.64 KB

README.md

File metadata and controls

78 lines (44 loc) · 1.64 KB
Raw

CTNet for Underwater Images

This is the deep learning algorithm developed for image matching in underwater environments.

Architecture of CTNet

CTNet

Environment

  • Python 3.8

  • PyTorch 1.13.0

  • TensorFlow 2.11.0

Dataset

The underwater images are from the dataset Tasmania Coral Point Count published by ACFR.

After downloading the Tasmania Coral Point Count dataset, reconstruct it as structure below:

Dataset
	- Group 1
		- Scene 1 Left
		- Scene 1 Right
	- Group 2
		...
	...
...

How to train

  1. Edit Hyper Parameter in option.py

  2. Run train.py

  3. All data will be saved under logs

How to test

  1. Load weights in option.py

  2. Run predict.py

  3. Insert the first picture

  4. Insert the second picture

  5. Obtain the similarity score

Citation

If you want to use this open source code for CTNet, please cite this github link.

Acknowledgement

Thanks for these open source publishers !!!

The code for CTNet is based on:

Siamese-PyTorch

ACT

The code for Siamese Network is based on:

Siamese-PyTorch

The code for MAML is based on:

MAML-PyTorch

The code for Reptile is based on:

Reptile-PyTorch