General Information

• Purpose: Classifier for a variety of concepts, common objects, etc. This model is a great all-purpose solution for most visual recognition needs with industry-leading performance.

• Architecture: Vision Transformer

• Intended Use: image indexing by tags, filtering, cascade routing

• Limitations: works well when content is prevalent in the image

Training/Test Data

The model was trained and tested on an internal dataset with approximately 10,000 concepts and 20M images, with multiple concepts per image. The class distributions on train and validation sets are long-tailed.