Datasets:

BrailleLens
/

Braille_Fingertip_Dataset

width int64 3.05k 4.05k	height int64 4.46k 5.88k	objects dict
3,844	5,505	{ "bbox": [ [ 588.7, 2066.1, 245.5, 224.2 ] ], "category": [ 0 ] }
3,872	5,390	{ "bbox": [ [ 2205.6, 1271.5, 256.9, 229.4 ] ], "category": [ 0 ] }
3,614	5,186	{ "bbox": [ [ 1501.9, 2030.2, 235.9, 187.2 ] ], "category": [ 0 ] }
3,844	5,548	{ "bbox": [ [ 1273.1, 297.9, 232.6, 202.2 ] ], "category": [ 0 ] }
3,746	5,407	{ "bbox": [ [ 2106.3, 4614.1, 248.5, 203 ] ], "category": [ 0 ] }
3,862	5,646	{ "bbox": [ [ 561.3, 2386.6, 275.8, 218.2 ] ], "category": [ 0 ] }
3,807	5,620	{ "bbox": [ [ 2012.6, 1869.1, 263.6, 190.9 ] ], "category": [ 0 ] }
3,860	5,761	{ "bbox": [ [ 920.3, 674, 258.1, 216.1 ] ], "category": [ 0 ] }
3,643	5,167	{ "bbox": [ [ 2248.2, 3942.8, 232.6, 209.3 ] ], "category": [ 0 ] }
3,983	5,513	{ "bbox": [ [ 1801.3, 2351.6, 253.7, 204.9 ] ], "category": [ 0 ] }
4,041	5,656	{ "bbox": [ [ 231.5, 4932.9, 265.9, 224.4 ] ], "category": [ 0 ] }
3,906	5,317	{ "bbox": [ [ 488.9, 432.9, 225.6, 176.9 ] ], "category": [ 0 ] }
3,794	5,486	{ "bbox": [ [ 2726.7, 507.9, 237.8, 175.7 ] ], "category": [ 0 ] }
3,745	5,448	{ "bbox": [ [ 745.2, 3187.6, 227.3, 193.9 ] ], "category": [ 0 ] }
3,793	5,485	{ "bbox": [ [ 2196.5, 3629, 229.7, 208.1 ] ], "category": [ 0 ] }
3,647	5,223	{ "bbox": [ [ 1413, 1391.7, 216.3, 148.8 ] ], "category": [ 0 ] }
4,022	5,750	{ "bbox": [ [ 362.4, 1467.9, 251.4, 188.6 ] ], "category": [ 0 ] }
4,052	5,792	{ "bbox": [ [ 1365.7, 3875.5, 241, 200 ] ], "category": [ 0 ] }
3,685	5,350	{ "bbox": [ [ 699.6, 707.9, 228.6, 177.1 ] ], "category": [ 0 ] }
4,048	5,876	{ "bbox": [ [ 674, 2345, 241.9, 209.3 ] ], "category": [ 0 ] }
3,910	5,665	{ "bbox": [ [ 738.7, 418.4, 217.4, 173.9 ] ], "category": [ 0 ] }
3,897	5,697	{ "bbox": [ [ 1219.2, 4513.1, 231.7, 219.5 ] ], "category": [ 0 ] }
3,681	5,290	{ "bbox": [ [ 1717.6, 2273.6, 214.3, 177.1 ] ], "category": [ 0 ] }
3,532	5,068	{ "bbox": [ [ 2773, 4085.2, 225.6, 181.4 ] ], "category": [ 0 ] }
3,559	5,126	{ "bbox": [ [ 851.6, 1885.1, 216.3, 179.1 ] ], "category": [ 0 ] }
3,490	5,086	{ "bbox": [ [ 421.7, 4511.6, 220.9, 186 ] ], "category": [ 0 ] }
3,544	5,076	{ "bbox": [ [ 869.1, 796.6, 217.1, 168.6 ] ], "category": [ 0 ] }
3,451	4,845	{ "bbox": [ [ 3144.4, 3353.6, 224.3, 156.8 ] ], "category": [ 0 ] }
3,682	5,184	{ "bbox": [ [ 1431.7, 844.3, 207, 169.8 ] ], "category": [ 0 ] }
3,688	5,458	{ "bbox": [ [ 2173.7, 3510.1, 221.6, 175.7 ] ], "category": [ 0 ] }
3,573	5,097	{ "bbox": [ [ 3245.5, 4948.5, 207.7, 141 ] ], "category": [ 0 ] }
3,379	4,925	{ "bbox": [ [ 382.2, 506.4, 202.4, 158.5 ] ], "category": [ 0 ] }
3,568	5,090	{ "bbox": [ [ 1819.9, 3786, 217.9, 148.7 ] ], "category": [ 0 ] }
3,345	4,856	{ "bbox": [ [ 7.6, 1756.4, 178.4, 175.7 ] ], "category": [ 0 ] }
3,433	4,974	{ "bbox": [ [ 2000.3, 869.4, 202.7, 164.9 ] ], "category": [ 0 ] }
3,264	4,748	{ "bbox": [ [ 208, 127.1, 172.9, 129.2 ] ], "category": [ 0 ] }
3,709	5,497	{ "bbox": [ [ 884.2, 1662, 200, 189.2 ] ], "category": [ 0 ] }
3,922	5,595	{ "bbox": [ [ 1270.1, 193, 206.1, 160.6 ] ], "category": [ 0 ] }
3,736	5,488	{ "bbox": [ [ 1892.3, 3683.2, 200, 167.6 ] ], "category": [ 0 ] }
3,392	4,849	{ "bbox": [ [ 800.9, 1533, 180.5, 146.3 ] ], "category": [ 0 ] }
3,431	5,001	{ "bbox": [ [ 2666.6, 2750.5, 206.5, 158.7 ] ], "category": [ 0 ] }
3,450	4,894	{ "bbox": [ [ 934.1, 562.2, 190.9, 160.6 ] ], "category": [ 0 ] }
3,593	4,996	{ "bbox": [ [ 910, 2834.5, 189.2, 162.2 ] ], "category": [ 0 ] }
3,284	4,703	{ "bbox": [ [ 642, 4292.2, 183.7, 151.2 ] ], "category": [ 0 ] }
3,291	4,659	{ "bbox": [ [ 1031.5, 1087.6, 176.7, 158.1 ] ], "category": [ 0 ] }
3,462	5,065	{ "bbox": [ [ 1214.7, 1767.4, 197.7, 155.8 ] ], "category": [ 0 ] }
3,189	4,615	{ "bbox": [ [ 1446.4, 317.9, 184.9, 150.9 ] ], "category": [ 0 ] }
3,054	4,459	{ "bbox": [ [ 1079.9, 1535.9, 181.4, 140 ] ], "category": [ 0 ] }

Braille Fingertip Dataset (YOLO format)

60 photographs of a reader's index finger resting on embossed Sinhala Braille pages, each annotated with one bounding box around the fingertip pad.

The dataset was created for BrailleLens, an offline Sinhala Braille learning app that runs on a smartphone camera or AiSee smart glasses. The app needs to know which Braille cell the learner is touching. Public fingertip datasets (TI1K, Roboflow) and MediaPipe Hands perform poorly in top-down Braille-reading views, where often only the fingertip is visible and the palm is out of frame. This set adapts fingertip detectors to that domain.

Dataset structure

images/
  train/   48 images + metadata.jsonl
  val/      6 images + metadata.jsonl
  test/     6 images + metadata.jsonl
labels/
  train/   48 YOLO .txt files
  val/      6 YOLO .txt files
  test/     6 YOLO .txt files
data.yaml  Ultralytics dataset config

Split	Images	Boxes
train	48	48
val	6	6
test	6	6

Splits are random (seed 42). Every image contains exactly one annotated fingertip.

Annotations

One class: 0: fingertip.

YOLO labels (labels/<split>/<image>.txt), one line per box, normalised to 0–1:

class x_center y_center width height
0 0.204330 0.389312 0.062350 0.041388

Hugging Face metadata (images/<split>/metadata.jsonl) holds the same boxes in pixel COCO format ([x_min, y_min, width, height]) for the dataset viewer and the datasets library:

{"file_name": "IMG_E6088.JPG", "width": 4118, "height": 5616,
 "objects": {"bbox": [[713.1, 2070.2, 256.8, 232.4]], "category": [0]}}

Annotation guideline: a tight rectangle around the distal pad of the reading finger (from the last knuckle through the nail), where skin meets the Braille page. Wrist and other fingers are excluded. Boxes were drawn in LabelMe and converted to YOLO format.

Collection

Device: Apple iPhone 14 Pro, 4118 × 5616 px JPEG
Content: a reader's index finger on embossed Sinhala Braille book pages, indoor lighting
Metadata: images contain no GPS location

Usage

Ultralytics YOLO

from huggingface_hub import snapshot_download
from ultralytics import YOLO

root = snapshot_download("BrailleLens/Braille_Fingertip_Dataset", repo_type="dataset")
model = YOLO("yolo26n.pt")
model.train(data=f"{root}/data.yaml", imgsz=640, epochs=50)

Hugging Face `datasets`

from datasets import load_dataset

ds = load_dataset("BrailleLens/Braille_Fingertip_Dataset")
sample = ds["train"][0]
sample["image"], sample["objects"]["bbox"]

Downloads last month: 243

Number of rows:

Total file size:

167 MB