ObjectNet

ImagesCreative Commons Attribution 4.0Introduced 2019-01-01

ObjectNet is a test set of images collected directly using crowd-sourcing. ObjectNet is unique as the objects are captured at unusual poses in cluttered, natural scenes, which can severely degrade recognition performance. There are 50,000 images in the test set which controls for rotation, background and viewpoint. There are 313 object classes with 113 overlapping ImageNet.

Source: On Robustness and Transferability of Convolutional Neural Networks Image Source: https://objectnet.dev/

Benchmarks

Image Classification/Top-1 Accuracy Image Classification/Top-5 Accuracy Image Classification/ARI Image Classification/Accuracy (%)Zero-Shot Transfer Image Classification/Accuracy (Private)Zero-Shot Transfer Image Classification/Accuracy (Public)Zero-Shot Transfer Image Classification/Top 5 Accuracy

Related Benchmarks

ObjectNet (All classes)/Object Recognition/Top 1 Accuracy ObjectNet (All classes)/Object Recognition/Top 5 Accuracy ObjectNet (Bounding Box)/Image Classification/Top 5 Accuracy ObjectNet (ImageNet classes)/Image Classification/Top 1 Accuracy ObjectNet (ImageNet classes)/Object Recognition/Top 1 Accuracy ObjectNet (ImageNet classes)/Object Recognition/Top 5 Accuracy ObjectNet (ImageNet classes, trained on ImageNet)/Object Recognition/Top 1 Accuracy ObjectNet (ImageNet classes, trained on ImageNet)/Object Recognition/Top 5 Accuracy