Detect out-of-distribution objects by size and border conditions
Box-conditional Hypothesis Testing for Out-of-Distribution Object Detection
Computer Vision and Pattern Recognition
Summary
Detecting objects that do not belong to known categories is important for safety in computer vision. The authors treat each detected object as a hypothesis test to decide if it belongs to a known class, adjusting their test based on object size and position in the image. They propose PLUM, a simple way to score how unfamiliar an object is by comparing it to known objects of similar size and position, using intermediate features of the detection model. Their method works well on standard benchmarks without needing extra models.
What this means in practice
- •For autonomous vehicle developers: Improve safety by reliably identifying unknown objects outside the training categories in vehicle perception systems.$Commercial implications: Enables safer autonomous driving systems by detecting unexpected objects on roads, which is critical for liability and passenger safety.
- •For security camera integrators: Enhance monitoring by flagging anomalous or unrecognized objects in surveillance footage without extra models.
Authors
Donghoon Lee, Jusung Park, Shinjin Kang
Abstract
Detecting out-of-distribution (OOD) objects is essential for ensuring the reliability and safety of object detectors in real-world deployment. We treat the decision on a detected box as a hypothesis test with the null hypothesis that the box covers an object of an in-distribution (ID) class. The null distribution changes with the size of the box and its contact with the image border, and the detector observes both. The test can therefore rank a box only against ID boxes of the same size and border contact, which gives a distribution-free bound on the false-alarm rate for each of them and uncovers the separation that one null distribution hides. The power of the test also rests on the layer it reads. Layers built only for the ID classes have no reason to keep what sets an unknown object apart, so the test reads the features before them. We propose \textbf{PLUM}, a training-free score that pools the last backbone stage inside the box and ranks its distance to the ID classes among ID boxes of the same size and border contact. On the standard benchmark, PLUM reaches the highest AUROC among the compared methods without external models, with 91.5 and 94.5 on PASCAL-VOC and 95.1 and 97.5 on BDD100K. Code: https://github.com/DDOZMI/PLUM