Datasets

Public data for pre-training

Create a new project from any of these

HierText-str

Image crops of generic text in the wild (with 4 to 32 tokens).

training: 17’092
validation: 3’496

scene text recognition

StanfordCars

Car make/model/year classification.

training: 8’144
validation: 8’041

multiclass classification

SKU110K

Grocery shelves full of products.

training: 8’219
validation: 588

quadrilateral detection

More coming soon!

Sihl AI