Datasets:

Formats:
arrow
ArXiv:
Libraries:
Datasets
SushantGautam commited on
Commit
ad0520d
·
verified ·
1 Parent(s): cc2265c

Update README.md

Browse files
Files changed (1) hide show
  1. README.md +112 -0
README.md CHANGED
@@ -29,3 +29,115 @@ dataset_info:
29
  download_size: 0
30
  dataset_size: 786835439
31
  ---
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
29
  download_size: 0
30
  dataset_size: 786835439
31
  ---
32
+
33
+ # 🩺 MedMultiPoints: A Multimodal Dataset for Object Detection, Localization, and Counting in Medical Imaging
34
+
35
+ [![Paper](https://img.shields.io/badge/CBMS-2025-blue)](https://arxiv.org/abs/2505.16647)
36
+ [![License: CC BY 4.0](https://img.shields.io/badge/License-CC%20BY%204.0-lightgrey.svg)](https://creativecommons.org/licenses/by/4.0/)
37
+ 📫 For queries, contact: [[email protected]](mailto:[email protected])
38
+
39
+ ---
40
+
41
+ ## Dataset Summary
42
+
43
+ **MedMultiPoints** is a curated, multimodal medical imaging dataset designed for **multi-task learning** in the medical domain—spanning **object detection**, **localization**, and **counting** tasks. It integrates data from **endoscopic** and **microscopic** modalities, reflecting real-world clinical diversity.
44
+
45
+ The dataset is introduced in the paper:
46
+ **"Point, Detect, Count: Multi-Task Medical Image Understanding with Instruction-Tuned Vision-Language Models"**
47
+ Presented at **IEEE CBMS 2025, Madrid, Spain.**
48
+ → [Project Page & Code](https://github.com/Simula/PointDetectCount)
49
+
50
+ ---
51
+
52
+ ## Features
53
+
54
+ - **10,600 images** from diverse modalities: endoscopy (HyperKvasir) and microscopy (VISEM-Tracking)
55
+ - Rich **multi-type annotations**:
56
+ - **Bounding Boxes** (`bbox_2d`) for object detection
57
+ - **Point Annotations** (`point_2d`) for localization
58
+ - **Count Labels** (`counts`) for counting tasks
59
+ - Compatible with **Vision-Language Models (VLMs)** and **instruction-tuned pipelines**
60
+ - JSON-formatted annotations designed for seamless integration with multimodal training
61
+
62
+ ---
63
+
64
+ ## Data Schema
65
+
66
+ Each sample in the dataset contains:
67
+
68
+ | Field | Type | Description |
69
+ |-------------------|-----------|--------------------------------------------------|
70
+ | `image` | Image | Raw medical image |
71
+ | `image_sha256` | string | SHA-256 hash of the image for integrity |
72
+ | `img_size` | [int, int]| Original image width and height |
73
+ | `points` | list | List of `[x, y]` point annotations |
74
+ | `bbox` | list | List of `[x1, y1, x2, y2]` bounding boxes |
75
+ | `count` | int | Object count in the image |
76
+ | `label` | string | Class label (e.g., `polyps`, `sperm`, etc.) |
77
+ | `collection_method` | string | Task type: `counting`, `detection`, etc. |
78
+ | `classification` | string | Description of annotation type (e.g., pathological-findings) |
79
+ | `organ` | string | Target organ: `Lower GI`, `Microscopy`, etc. |
80
+
81
+ ---
82
+
83
+ ## Supported Tasks
84
+
85
+ This dataset supports the following **multi-task** settings:
86
+
87
+ - 🔲 **Object Detection** (bounding box prediction)
88
+ - 📍 **Localization** (point prediction)
89
+ - 🔢 **Counting** (object count regression)
90
+ - 🧠 **Multimodal Instruction-Based Learning**
91
+
92
+ ---
93
+
94
+ ## How to Load
95
+
96
+ ```python
97
+ from datasets import load_dataset
98
+
99
+ ds = load_dataset("SushantGautam/MedMultiPoints")['train']
100
+ sample = ds[0]
101
+
102
+ # Access image and annotations
103
+ image = sample['image']
104
+ bbox = sample['bbox']
105
+ points = sample['points']
106
+ count = sample['count']
107
+ ```
108
+
109
+ ---
110
+
111
+ ## Example
112
+
113
+ ```json
114
+ {
115
+ "image_sha256": "71179abc4b011cc99bddb3344e3e114765b32bdf77e78892f046026d785a4bdb",
116
+ "img_size": [622, 529],
117
+ "points": [[234, 171.5]],
118
+ "bbox": [[38, 5, 430, 338]],
119
+ "count": 1,
120
+ "label": "polyps",
121
+ "collection_method": "counting",
122
+ "classification": "pathological-findings",
123
+ "organ": "Lower GI"
124
+ }
125
+ ```
126
+
127
+ ---
128
+
129
+ ## Citation
130
+
131
+ If you use this dataset, please cite:
132
+
133
+ ```bibtex
134
+ @article{Gautam2025May,
135
+ author = {Gautam, Sushant and Riegler, Michael A. and Halvorsen, P{\aa}l},
136
+ title = {{Point, Detect, Count: Multi-Task Medical Image Understanding with Instruction-Tuned Vision-Language Models}},
137
+ journal = {arXiv},
138
+ year = {2025},
139
+ month = may,
140
+ eprint = {2505.16647},
141
+ doi = {10.48550/arXiv.2505.16647}
142
+ }
143
+ ```