DGTRSD

# DGTRSD Dataset This repository hosts the split archive files of **DGTRSD**, the Dual-Granularity Remote Sensing image-text dataset used by DGTRS-CLIP. Code repository: [MitsuiChen14/DGTRS](https://github.com/MitsuiChen14/DGTRS) ## Files The dataset archive is split into 5GB chunks named `DGTRSD_part.tar.gz.*`. To reconstruct the original archive after downloading all chunks: ```bash cat DGTRSD_part.tar.gz.* > DGTRSD.tar.gz tar -xzf DGTRSD.tar.gz ``` ## Caption JSON Files The `caption_json/` folder contains dual-granularity caption annotation files used by DGTRS-CLIP training: - `caption_json/MillionAID.json` - `caption_json/crowdai.json` - `caption_json/BigEarthNet.json` - `caption_json/loveda.json` - `caption_json/cvact.json` - `caption_json/cvusa.json` - `caption_json/fmow-big-bbox-center.json` - `caption_json/fmow-small.json` - `caption_json/Det-10.json` - `caption_json/Seg-4.json` - `caption_json/UCMerced.json` - `caption_json/OpenLandMap.json`