NAO

Natural Adversarial Object

Dataset Information
Modalities
Images
Introduced
2021
License
None
Homepage

Overview

Natural Adversarial Objects (NAO) is a new dataset to evaluate the robustness of object detection models. NAO contains 7,934 images and 9,943 objects that are unmodified and representative of real-world scenarios, but cause state-of-the-art detection models to misclassify with high confidence.

Variants: NAO

Associated Benchmarks

This dataset is used in 1 benchmark:

Recent Benchmark Submissions

Task Model Paper Date
Object Detection RetinaNet-R50 Natural Adversarial Objects 2021-11-07
Object Detection YOLOv3 Natural Adversarial Objects 2021-11-07
Object Detection EfficientDet-D2 Natural Adversarial Objects 2021-11-07
Object Detection Mask RCNN R50 Natural Adversarial Objects 2021-11-07
Object Detection EfficientDet-D4 Natural Adversarial Objects 2021-11-07
Object Detection EfficientDet-D7 Natural Adversarial Objects 2021-11-07
Object Detection Faster RCNN Natural Adversarial Objects 2021-11-07

Research Papers

Recent papers with results on this dataset: