This dataset provides the cleaned text captions (.clean.txt) used for training the GeoMTL model on satellite image captioning and Visual Question Answering (VQA) tasks for agricultural earth observation.
This repository contains only the language targets (text annotations) generated for the Multi-Temporal Crop Classification dataset. To fully utilize this data for multi-task learning (segmentation… See the full description on the dataset page:
https://huggingface.co/datasets/trust-tad/geomtl-dataset.
