STHN — Two-Stages Model

This is the two-stages variant of STHN (Satellite-Thermal Homography Network) for UAV satellite-thermal geo-localization. It performs coarse homography estimation followed by a fine-stage refinement on a cropped region around the coarse prediction.

Pushed to the Hub using the PytorchModelHubMixin integration.

Config

{
  "model_config": {
    "corr_level": 4,
    "database_size": 1536,
    "fine_padding": 0,
    "iters_lev0": 6,
    "iters_lev1": 6,
    "resize_width": 256,
    "two_stages": true
  }
}
Downloads last month
34
Safetensors
Model size
3.06M params
Tensor type
F32
·
Inference Providers NEW
This model isn't deployed by any Inference Provider. 🙋 Ask for provider support

Dataset used to train xjh19972/STHN-two-stages

Collection including xjh19972/STHN-two-stages

Paper for xjh19972/STHN-two-stages