CoolFace
Datasetpublic

Panorama-grounding/MRSeg-Referring-Expressions

MRSeg Referring Expressions (single-turn) Single-turn referring-expression segmentation samples derived from the multi-round MR-Seg data of SegLLM (paper). SegLLM's original conversations provide the target regions as input to later rounds, which makes them unusable for models that receive only an image and text. We used Qwen3-VL-32B-Instruct to keep only the samples whose target can be identified from the image and expression alone, rewriting each into a self-contained… See the full description on the dataset page: https://huggingface.co/datasets/Panorama-grounding/MRSeg-Referring-Expressions.

sourceHugging Facecc-by-nc-4.0updated 9d agoView on Hugging Face
2likes179downloads
Dataset Card

MRSeg Referring Expressions (single-turn)

![Paper](https://arxiv.org/abs/2609.19143) ![Code](https://github.com/sarapieri/panorama_grounding) ![Project](https://huggingface.co/Panorama-grounding) ![Website](https://www.di.ens.fr/willow/research/panorama/)

Single-turn referring-expression segmentation samples derived from the multi-round MR-Seg data of SegLLM (paper).

SegLLM's original conversations provide the target regions as input to later rounds, which makes them unusable for models that receive only an image and text. We used Qwen3-VL-32B-Instruct to keep only the samples whose target can be identified from the image and expression alone, rewriting each into a self-contained referring expression. The segmentation masks are resolved from the source datasets' annotations and stored inline as COCO RLE, so no external annotation files are needed — only the images.

The release contains 101,349 samples spanning parts, stuff regions, attributes, and relationally specified regions. This is the multi-granularity bucket of PANORAMA's training mixture.

See also our human-annotated benchmark for panoptic grounded captioning, PanoCaps, and our re-captioned COCONut-PanCap.

What is in this repository

PathContents
data/train-*.parquetone row per sample, mask inline as COCO RLE
mrseg_accepted.tarthe same samples as per-source .jsonl shards, as read by our loader

Download

bash
hf download Panorama-grounding/MRSeg-Referring-Expressions --repo-type dataset \
    --local-dir /path/to/data/MRSeg-Referring-Expressions
mkdir -p /path/to/data/mrseg_extract && tar -xf /path/to/data/MRSeg-Referring-Expressions/mrseg_accepted.tar \
    -C /path/to/data/mrseg_extract   # accepted/*.jsonl

Images

Images are not included. Samples reference them by a relative path (img_rel) under four public sources:

`img_rel` prefixDatasetDownload
coco/COCO train2017cocodataset.org
ade20k/ADE20KMIT Scene Parsing
pascal/PASCAL VOC 2010host.robots.ox.ac.uk
VG_100K/, VG_100K_2/Visual Genomevisualgenome.org

LVIS, PACO and COCO-Stuff annotate COCO images, so they require no additional image download.

Fields

FieldDescription
sidUnique sample id, e.g. ade20k_016857_00.
sourceSource bucket: mr_lvis, mr_paco, mr_pascal, mr_vg, ade20k, cocostuff or attributes.
img_relImage path relative to the source dataset root.
expressionSelf-contained referring expression, e.g. "the stone stairs in the colonnade".
instructionA natural-language segmentation instruction wrapping the expression (instr in the jsonl).
maskCOCO RLE (size, counts) in the jsonl; flattened to mask_size / mask_counts in the parquet.
model, prompt_verExtraction provenance (jsonl only).

License

This dataset is derived from SegLLM's MR-Seg data and from the annotations of LVIS, PACO, Visual Genome, COCO-Stuff, ADE20K and PASCAL-Part; it follows the licenses of these source datasets. For research and non-commercial use only.

Citation

If you use these expressions, please cite our paper and the original SegLLM work:

bibtex
@article{pieri2026panorama,
  title   = {PANORAMA: Panoptic Grounded Captioning via Mask Proposal Selection},
  author  = {Pieri, Sara and Kazakos, Evangelos and Chen, Shizhe and Sivic, Josef and Schmid, Cordelia},
  journal = {arXiv preprint arXiv:2609.19143},
  year    = {2026}
}

@inproceedings{wang2025segllm,
  title     = {{SegLLM}: Multi-round Reasoning Segmentation with Large Language Models},
  author    = {Wang, XuDong and Zhang, Shaolun and Li, Shufan and Kallidromitis, Konstantinos and Li, Kehan and Kato, Yusuke and Kozuka, Kazuki and Darrell, Trevor},
  booktitle = {International Conference on Learning Representations},
  year      = {2025}
}