Open Science

12 research datasets.

AIRL believes benchmark data belongs to the community. Four collections are fully open; the rest are shared with researchers on request.

Publicly available

Download today

ArASL Arabic sign language dataset sample
ArASL — Arabic Alphabets Sign Language. 54,049 fully labelled images of 32 signs from 40 participants. Mendeley Data · Data in Brief (2019).
Arabic traffic sign dataset sample
ArTS — Arabic Traffic Signs. 2,718 real + 57,078 augmented images across 24 sign classes. Mendeley Data (2020).
Writer verification dataset sample
Arabic Writer-Verification Characters. 10,780 characters extracted from handwritten Arabic text. PeerJ CS article + data (2022).

DeepFruit

41,509 labelled fruit/vegetable images across 16 classes for classification and calorie estimation. Published with Data in Brief (2023).

Available on request

Write to us for access

Students' Grades

250-student, three-semester dataset for "at-risk" prediction research.

Arabic Handwritten Numerals

Benchmark numerals for multi-language recognition.

Arabic Handwritten Words

Word-level handwriting corpus.

Arabic Braille

Braille character & word imagery behind the patented reader.

Arabic Hate Speech

Labelled social-network text for offense detection.

Microscopic Mineral Grains

SEM-annotated grain imagery from the FRQNT program.

Date Fruits (10 classes)

Variety-classification imagery.

Date-Fruit Diseases (4 classes)

Disease-recognition imagery.

Request access: glatif@tru.ca