objektenet
ObjectNet is a dataset designed to evaluate the robustness of object recognition models. It was created by researchers at MIT and consists of over 100,000 images featuring various object categories. The key characteristic of ObjectNet is that it presents objects in challenging and often unnatural contexts. This includes scenarios where objects are rotated, zoomed, occluded, or appear in unusual backgrounds and viewpoints. The goal is to assess whether models trained on standard datasets can generalize to these more difficult real-world situations.
The dataset was curated to include images that are likely to cause errors in current computer vision