Visual Object Retrieval in satellite images with Triplet Network generated from Template (Processing)
HydroViet_VOR
|
└───train_data
│ └───house
│ └───house1.jpg
│ └───house2.jpg
│ └───...
│ └───river
│ └───...
| └───...
|
└───val_data
│ └───house
│ └───house1.jpg
│ └───house2.jpg
│ └───...
│ └───river
│ └───...
| └───...
|
└───test
| │ 1.jpg
| │ 2.jpg
| | ...
python trainv1.py
python test.py
python visualize.py
The best accuracy when validating is 56%, which is not good enough
Using GANs to train CNN models like EfficientNet, Resnet on satellite data
Make sure to give them a star!