Woodscape Fisheye Semantic Segmentation for Autonomous Driving -- CVPR 2021 OmniCV Workshop Challenge

We present the WoodScape fisheye semantic segmentation challenge for\nautonomous driving which was held as part of the CVPR 2021 Workshop on\nOmnidirectional Computer Vision (OmniCV). This challenge is one of the first\nopportunities for the research community to evaluate the semantic segmentation\ntechniques targeted for fisheye camera perception. Due to strong radial\ndistortion standard models don't generalize well to fisheye images and hence\nthe deformations in the visual appearance of objects and entities needs to be\nencoded implicitly or as explicit knowledge. This challenge served as a medium\nto investigate the challenges and new methodologies to handle the complexities\nwith perception on fisheye images. The challenge was hosted on CodaLab and used\nthe recently released WoodScape dataset comprising of 10k samples. In this\npaper, we provide a summary of the competition which attracted the\nparticipation of 71 global teams and a total of 395 submissions. The top teams\nrecorded significantly improved mean IoU and accuracy scores over the baseline\nPSPNet with ResNet-50 backbone. We summarize the methods of winning algorithms\nand analyze the failure cases. We conclude by providing future directions for\nthe research.\n