AwA2

Animals with Attributes 2

ImagesIntroduced 2019-01-01

Animals with Attributes 2 (AwA2) is a dataset for benchmarking transfer-learning algorithms, such as attribute base classification and zero-shot learning. AwA2 is a drop-in replacement of original Animals with Attributes (AwA) dataset, with more images released for each category. Specifically, AwA2 consists of in total 37322 images distributed in 50 animal categories. The AwA2 also provides a category-attribute matrix, which contains an 85-dim attribute vector (e.g., color, stripe, furry, size, and habitat) for each category.

Source: Learning from Noisy Web Data with Category-level Supervision Image Source: https://arxiv.org/pdf/1604.00326.pdf

Benchmarks

Concept-based Classification/Task Accuracy (%)Concept-based Classification/Concept Accuracy (%)Generalized Few-Shot Learning/Per-Class Accuracy (1-shot)Generalized Few-Shot Learning/Per-Class Accuracy (2-shots)Generalized Few-Shot Learning/Per-Class Accuracy (5-shots)Generalized Few-Shot Learning/Per-Class Accuracy (10-shots)Generalized Few-Shot Learning/Per-Class Accuracy (20-shots)Image Classification/Task Accuracy (%)Image Classification/Concept Accuracy (%)Zero-Shot Learning/average top-1 classification accuracy Zero-Shot Learning/Accuracy Seen Zero-Shot Learning/Accuracy Unseen Zero-Shot Learning/H Zero-Shot Learning/Harmonic mean

Related Benchmarks

AWA2 - 0-Shot/Few-Shot Image Classification/Accuracy AWA2 - 0-Shot/Image Classification/Accuracy