File size: 3,783 Bytes
42229cc
 
 
 
c14e918
509e03e
c14e918
42229cc
 
 
 
 
 
9582923
42229cc
 
 
 
 
 
 
 
 
 
 
 
ae480bf
 
 
42229cc
ae480bf
 
9befd4e
 
 
68b1470
ae480bf
 
 
 
 
 
 
 
 
 
 
 
 
 
 
680a0ca
 
 
 
 
 
 
 
 
 
 
 
 
 
 
ae480bf
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
---
license: cc-by-nc-4.0
---
# R2SM Dataset

<img src="https://cdn-uploads.huggingface.co/production/uploads/68231c1c344454ad3d607d95/nYzy-uI2YLR93ll1Bc3Mo.jpeg" alt="R2SM Dataset Image" width="600"/>

## Dataset preparation
### 1. COCOA-cls
Download COCO images (`2014 Train images`, `2014 Val images`) from [here](https://cocodataset.org/#download).
### 2. D2SA
Download D2SA images from [here](https://www.mvtec.com/company/research/datasets/mvtec-d2s/).
### 3. MUVA
Download MUVA images from [here](https://drive.google.com/drive/folders/1T5PNhoWlXBFDwGteVi3x357adM1t2mlo).

## Dataset format
Each split (`cocoa-cls_split`, `d2sa_split`, `muva_split`) follows the gRefCOCO format and contains three files:
1. `instances.json`
    * Contains all instance annotations
    * **Mask format**: RLE
    * **Bounding box format**: [x, y, width, height]
2. `queries_amodal.json`
    * Includes amodal queries only (as mentioned in the main paper).
    * Each entry links a query to the referred objects. 
5.  `queries_all.json`
    * Includes both amodal and modal queries.
    * Each entry links queries to the referred objects.
  
## Source Attribution

The R2SM dataset is constructed using images and annotations adapted from the following publicly available datasets:

- **COCOA-cls** and **D2SA**: From *Learning to See the Invisible: End-to-End Trainable Amodal Instance Segmentation (WACV 2019)* by Follmann et al.
- **MUVA**: From *MUVA: A New Large-Scale Benchmark for Multi-view Amodal Instance Segmentation in the Shopping Scenario (ICCV 2023)* by Li et al.
  - Licensed under Creative Commons Attribution-NonCommercial 4.0 International (CC BY-NC 4.0).
  - License link: https://creativecommons.org/licenses/by-nc/4.0/

All images and annotations are originally released under non-commercial academic licenses, and R2SM is released under the same usage restriction. 
Please refer to the original datasets for full details.

## Citations

If you use R2SM, please also cite the original sources:

```bibtex
@inproceedings{FollmannKHKB19,
  author       = {Patrick Follmann and Rebecca K{\"{o}}nig and Philipp H{\"{a}}rtinger and Michael Klostermann and Tobias B{\"{o}}ttger},
  title        = {Learning to See the Invisible: End-to-End Trainable Amodal Instance Segmentation},
  booktitle    = {{IEEE} Winter Conference on Applications of Computer Vision, {WACV} 2019, Waikoloa Village, HI, USA, January 7-11, 2019},
  year         = {2019}
}
@inproceedings{FollmannBHKU18,
  author       = {Patrick Follmann and Tobias B{\"{o}}ttger and Philipp H{\"{a}}rtinger and Rebecca K{\"{o}}nig and Markus Ulrich},
  editor       = {Vittorio Ferrari and Martial Hebert and Cristian Sminchisescu and Yair Weiss},
  title        = {MVTec {D2S:} Densely Segmented Supermarket Dataset},
  booktitle    = {Computer Vision - {ECCV} 2018 - 15th European Conference, Munich, Germany, September 8-14, 2018, Proceedings, Part {X}},
  series       = {Lecture Notes in Computer Science},
  year         = {2018}
}
@inproceedings{ZhuTMD17,
  author       = {Yan Zhu and Yuandong Tian and Dimitris N. Metaxas and Piotr Doll{\'{a}}r},
  title        = {Semantic Amodal Segmentation},
  booktitle    = {2017 {IEEE} Conference on Computer Vision and Pattern Recognition,
                  {CVPR} 2017, Honolulu, HI, USA, July 21-26, 2017},
  year         = {2017}
}
@inproceedings{LiYTBZJ023,
  author       = {Zhixuan Li and Weining Ye and Juan Terven and Zachary Bennett and Ying Zheng and Tingting Jiang and Tiejun Huang},
  title        = {{MUVA:} {A} New Large-Scale Benchmark for Multi-view Amodal Instance Segmentation in the Shopping Scenario},
  booktitle    = {{IEEE/CVF} International Conference on Computer Vision, {ICCV} 2023, Paris, France, October 1-6, 2023},
  year         = {2023}
}