search

AnkurDeria / MFT

Pytorch implementation of Multimodal Fusion Transformer for Remote Sensing Image Classification.
186 stars 19 forks source link

Classification map generation with both HSI and Lidar Data provided to MFT model for Trento #14

Open jilaniw opened 1 month ago

jilaniw commented 1 month ago

Hi, I am really grateful for the prompt responses for last issues. I tried to generate classification map with your model provided for Trento data. The classification map looks like below. I am unbale to understand why it's not same like shown in your transaction. The fig below shows map of few classes only, that is also not correct. As the model is multimodal and needs two data for prediction: complete HSI (including background pixels) as well as lidar. I hope you will help in this matter also.

classification_map_MF_Trento classification_map_MFT_minmax_spectral

AnkurDeria commented 1 month ago

Make sure the patch input is correct for both HSI and LiDAR, including the background pixels. The model doesn't output a background class, so the output image will only have 6 different colours, not 7.