README.md

---
dataset_info:
  features:
  - name: split
    dtype: string
  - name: image_id
    dtype: string
  - name: file_name
    dtype: string
  - name: image_info
    struct:
    - name: data_source
      dtype: string
    - name: file_name
      dtype: string
    - name: height
      dtype: int64
    - name: id
      dtype: string
    - name: width
      dtype: int64
  - name: caption_info
    struct:
    - name: caption
      dtype: string
    - name: caption_ann
      dtype: string
    - name: id
      dtype: int64
    - name: image_id
      dtype: string
    - name: label_matched
      list:
      - name: mask_ids
        sequence: int64
      - name: txt_desc
        dtype: string
    - name: labels
      sequence: string
  - name: mask_annotations
    list:
    - name: area
      dtype: int64
    - name: bbox
      sequence: float64
    - name: category_id
      dtype: int64
    - name: id
      dtype: int64
    - name: image_id
      dtype: string
    - name: iscrowd
      dtype: int64
    - name: segmentation
      struct:
      - name: counts
        dtype: string
      - name: size
        sequence: int64
    - name: thing_or_stuff
      dtype: string
  - name: categories
    list:
    - name: id
      dtype: int64
    - name: name
      dtype: string
  splits:
  - name: train
    num_bytes: 29443350
    num_examples: 2070
  - name: val
    num_bytes: 4782919
    num_examples: 420
  - name: test
    num_bytes: 10976834
    num_examples: 980
  download_size: 25273455
  dataset_size: 45203103
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: val
    path: data/val-*
  - split: test
    path: data/test-*
---
# PanoCaps (PANORAMA): Panoptic grounded captioning via mask-guided refinement

[![Paper](https://img.shields.io/badge/ArXiv-Paper-brown)]()
[![Code](https://img.shields.io/badge/GitHub-Link-orange)](https://github.com/sarapieri/panorama_grounding)
[![Data](https://img.shields.io/badge/HuggingFace-Data-blue)](https://huggingface.co/datasets/HuggingSara/PanoCaps)
[![Website](https://img.shields.io/badge/Web-Page-purple)](https://www.di.ens.fr/willow/research/panorama/)
<!-- [![Model](https://img.shields.io/badge/HuggingFace-Model-green)]() -->

<p align="center">
  <img src="https://www.di.ens.fr/willow/research/panorama/resources/panorama_teaser.jpg"
       width="100%"
       alt="Panorama teaser image" />
</p>

PanoCaps is a unified dataset for **panoptic grounded captioning**. A model must generate a full-scene caption and ground every mentioned entity (things and stuff) with pixel-level masks. 

Every caption:
- Is **human-written**
- Covers the **entire visible scene**  
- Contains **rich open-vocabulary descriptions** beyond category labels  
- Includes **inline grounding tags** referring to segmentation masks  
- Supports **one-to-many** and **many-to-one** text ↔ mask mappings  

This makes PanoCaps suitable for training and evaluating **vision–language models** requiring both detailed scene understanding and fine-grained spatial grounding.

The repository includes:

1. **Raw annotations** in JSON format (`annotations/`) → best for **training & evaluation**  
2. **A processed Hugging Face dataset**  → best for **visualization & inspection**

This dataset is intended **exclusively for research and non-commercial use**.

## Dataset Details

### Dataset Description

This benchmark supports **panoptic grounded captioning**—a task requiring models to generate long-form, descriptive captions for the entire scene and link all mentioned entities (things and stuff) to pixel-level masks. Masks follow standard **COCO-style panoptic annotations**.

The dataset comprises **3,470 images** with a total of **34K panoptic regions**, averaging **~9 grounded entities per image**. The human-written captions are designed for maximum quality and detail:
* **Comprehensive:** Covers the entire visible scene.
* **Open-Vocabulary:** Entity descriptions extend beyond simple category labels.
* **Fully Grounded:** Uses in-text markers and explicit mapping structures (`label_matched`) to link text spans to masks, ensuring **>99% of regions are grounded**.


### Images

**Images are *not* included** in this repository.

To use the dataset, download the original images from the source datasets:

| Dataset | Data Download Link                          | Associated Publication |
|---------|---------------------------------------------|-----------------------------------------|
| ADE20K  | [ADE20K Download](https://groups.csail.mit.edu/vision/datasets/ADE20K/) | [ADE20K Paper](https://arxiv.org/abs/1608.05442)                 |
| COCONut | [COCONut GitHub](https://github.com/bytedance/coconut_cvpr2024) | [COCONut Paper](https://arxiv.org/abs/2404.08639)                        |
| VIPSeg  | [VIPSeg GitHub](https://github.com/VIPSeg-Dataset/VIPSeg-Dataset/) | [VIPSeg Paper](https://openaccess.thecvf.com/content/CVPR2022/html/Miao_Large-Scale_Video_Panoptic_Segmentation_in_the_Wild_A_Benchmark_CVPR_2022_paper.html)                      |


The JSON annotations reference these images by consistent `file_name` and `id`.

### Repository Structure

<details>
    <summary>Show Repository Structure</summary>
    <pre>
    PanoCaps/
    │
    ├── 📁 annotations/
    │   ├── 📄 test_caption.json
    │   ├── 📄 test_mask.json
    │   ├── 📄 train_caption.json
    │   ├── 📄 train_mask.json
    │   ├── 📄 val_caption.json
    │   └── 📄 val_mask.json
    ├── 📁 data/ (parquet/HF version)
    └── 📄 README.md
    </pre>
</details>

### Recommended Usage

This dataset is provided in two complementary formats:

### **1. Hugging Face Dataset Format (recommended for inspection & visualization)**
The `train`, `val`, and `test` splits uploaded to the Hugging Face Hub combine **captioning** and **panoptic mask** information into a **single unified entry per image**. This format is ideal for browsing samples interactively in the Dataset Viewer or quick experimentation.

### **2. Original COCO-Style JSON Format (recommended for training & evaluation)**
Raw annotations are provided under `annotations/` as pairs of Caption files and Mask files (e.g., `train_caption.json` / `train_mask.json`).

These follow the original COCO-style structure and are best suited for:
- Model training  
- Model evaluation  
- Direct integration into COCO-based pipelines  

Caption and mask files can be matched using the shared `image_id` / `id` fields in `images[*]` and `annotations[*]`. 

### Detailed COCO Format

<details>
  <summary>Show Caption File Example (Structure + Single Entry)</summary>

```javascript
{
  "annotations": [
    {
      "caption": "The image shows a small, brightly lit bathroom dominated by a white tiled wall...",
      // Clean natural-language caption
      "caption_ann": "The image shows a small, brightly lit bathroom dominated by a <0:white tiled wall>...",
      // Caption with grounded <mask_id:text> references
      "label_matched": [
        { "mask_ids": [0], "txt_desc": "white tiled wall" },
        { "mask_ids": [5], "txt_desc": "white bathtub with chrome faucets" }
        // ...
      ],
      // Mapping text spans → one or more mask IDs
      // Masks may appear multiple times with different descriptions
      "id": 0,
      // Caption annotation ID
      "image_id": "00000006",
      // Matches the images[*].id field
      "labels": ["wall", "floor", "ceiling", "window", "curtain", "tub", "sink"]
      // All unique semantic labels from the original annotations
    }
  ],
  "images": [
    {
      "file_name": "00000006.jpg",
      // Image filename
      "height": 973,
      "width": 512,
      // Image resolution
      "id": "00000006",
      // Image identifier (matches annotation.image_id)
      "data_source": "ADE20K"
      // Image source
    }
  ]
}
```
</details>

<details>
  <summary>Show Mask File Example (Structure + Single Entry)</summary>

```javascript
{
  "annotations": [
    {
      "id": 0,
      // Unique ID of this panoptic region
      "image_id": "00000006",
      // Links this region to the image and caption (matches images[*].id and caption image_id)
      "category_id": 100,
      // Semantic category ID (from the original annotations)
      "segmentation": {
        "size": [973, 512],
        // Height and width of the full image (needed to decode the RLE mask)
        "counts": "d1`1Zk0P2C=C<D=C=C6J=..." 
        // RLE-encoded mask in COCO panoptic format
      },
      "area": 214858,
      // Number of pixels covered by this segment
      "bbox": [0.0, 0.0, 511.0, 760.0],
      // COCO-format bounding box [x, y, width, height]
      "iscrowd": 0,
      // 0 for normal segment, 1 if this region is a crowd
      "thing_or_stuff": "stuff"
      // Whether this region is an object-like "thing" or background-like "stuff"
    }
  ],
  "images": [
    {
      "file_name": "00000006.jpg",
      // Image file name (in the original dataset)
      "height": 973,
      "width": 512,
      // Image resolution
      "id": "00000006"
      // Image identifier (matches annotations[*].image_id and caption image_id)
      "data_source": "ADE20K"
      // Image source
    }
  ],
  "categories": [
    {
      "id": 1,
      // Category ID (referenced by annotations[*].category_id)
      "name": "object"
      // Human-readable category name
    }
  ]
}
```
</details>

---

## Curation and Annotation Details

PanoCaps was built to overcome the limitations of prior grounded captioning datasets (e.g., auto-generated captions, limited vocabulary, and incomplete grounding). Our goal was to create a resource where captions describe every meaningful region using open-vocabulary language, with explicit grounding for each referenced entity.
The creation process involved four stages:

1.  **Image Selection:** A diverse subset of images was curated from ADE20K, COCONut, and VIPSeg to ensure visual quality and suitability for dense grounding.
2.  **Captioning:** Professional annotators wrote long-form, fine-grained scene descriptions, highlighting attributes, relationships, and all visible entities.
3.  **Grounding:** Annotators tagged textual references with `<mask_id:description>` markers and produced **label_matched** structures that map text spans to one or more segmentation masks.
4.  **Validation:** A second QC stage verified the correctness of grounding IDs, completeness of region coverage, and annotation consistency.
**Data Producers:** The base panoptic masks were sourced from the original datasets (ADE20K, COCONut, VIPSeg). However, all **captions and grounding annotations** were created specifically for PanoCaps by paid professional annotators following internal guidelines.

---

## License (Research Only)
Because this repository merges, normalizes, and redistributes content from already existing datasets, the combined dataset is provided **strictly for research and non-commercial use**.
Commercial use is **not permitted**. Users must comply with the licenses of each original source dataset.

---

## Citation
If you find our work useful for your research, please consider citing our [paper]():
```
@article{YOUR_CITATION_HERE,
  title={Your Title},
  author={Your Name},
  year={2024}
}
```