Identifying unexpected objects on roads in semantic segmentation is crucial in safety-critical applications . Existing approaches use images of unexpected objects from external datasets or require additional training . We propose a simple yet effective approach that standardizes the max logits in order to align the different distributions and reflect the relative meanings of each predicted class . Such a straightforward approach achieves a new state-of-the-artperformance on the publicly available Fishyscapes Lost & Found leaderboard with a large margin . In contrast to previous approaches, ourmethod does not utilize any external datasets . It makes our method widely applicable to existing pre-trained segmentation models . We consider the local regionsfrom two different perspectives based on the intuition that neighboring pixelsshare similar semantic information . We also consider the Local Regions from two different . perspectives . We hope to use the Fishyscape leaderboard to help identify unexpected objects in the leaderboard by the end of the leaderboards witha large margin. In contrast,

Author(s) : Sanghun Jung, Jungsoo Lee, Daehoon Gwak, Sungha Choi, Jaegul Choo

Links : PDF - Abstract

Code :
Coursera

Keywords : unexpected - objects - segmentation - leaderboard - approach -

Leave a Reply

Your email address will not be published. Required fields are marked *