search

ViTAE-Transformer / MTP

The official repo for [JSTARS'24] "MTP: Advancing Remote Sensing Foundation Model via Multi-Task Pretraining"
MIT License
176 stars 11 forks source link

Inference on single image #28

Closed ogencoglu closed 3 weeks ago

ogencoglu commented 1 month ago

Is there a quick script, notebook or colab to run semantic segmentation inference with your pretrained models models on some arbitrary image that is not part of the datasets you listed?

DotWang commented 3 weeks ago

@ogencoglu may be you can find it in the SAMRS repo

ogencoglu commented 3 weeks ago

@DotWang what do you mean maybe? You are the main developer of SAMRS. Is it there or not?

DotWang commented 3 weeks ago

@ogencoglu What you need is an ordinary function in previous research about segmentation, while the sliding window inference for any size image is achieved by the repo of PSPNet in a very early time. Even if my code is originally borrowed from there, so all of my repos related to segmentation (except openmmlab-based) may contain it, such as samrs, dcn-t, etc.

DotWang commented 3 weeks ago

@ogencoglu maybe this your need

https://github.com/ViTAE-Transformer/SAMRS/blob/main/Pretraining%20and%20Finetuning/Encoder_Decoder/test_gpu.py

or you can use the mmsegmentation

ogencoglu commented 3 weeks ago

Thanks!