Ashish Patel 🇮🇳’s Post

𝗗𝗮𝘆-𝟯𝟵𝟳 𝗖𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗧𝗵𝗲 𝗞𝗙𝗜𝗼𝗨 𝗟𝗼𝘀𝘀 𝗳𝗼𝗿 𝗥𝗼𝘁𝗮𝘁𝗲𝗱 𝗢𝗯𝗷𝗲𝗰𝘁 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗯𝘆 𝗔𝗜 𝗜𝗻𝘀𝘁𝗶𝘁𝘂𝘁𝗲, 𝗦𝗵𝗮𝗻𝗴𝗵𝗮𝗶 𝗝𝗶𝗮𝗼 𝗧𝗼𝗻𝗴 𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗶𝘁𝘆  Follow me for a similar post: Ashish Patel  ------------------------------------------------------------------- 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗙𝗮𝗰𝘁𝘀 : 🔸 Paper: 𝗧𝗵𝗲 𝗞𝗙𝗜𝗼𝗨 𝗟𝗼𝘀𝘀 𝗳𝗼𝗿 𝗥𝗼𝘁𝗮𝘁𝗲𝗱 𝗢𝗯𝗷𝗲𝗰𝘁 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 🔸 This paper is published arxiv2022. 🔸 Presented a trend-level consistent approximate to the ideal but gradient-training unfriendly SkewIoU loss for rotation detection, and we call it KFIoU loss as the Kalman filter is adopted to directly mimic the computing mechanism of SkewIoU by definition. This design differs from the distribution distance-based losses including GWD and KLD which in our analysis have fundamental difficulty in achieving trend-level alignment with SkewIoU loss. ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 Differing from the well-developed horizontal object detection area whereby the computing-friendly IoU based loss is readily adopted and well fits with the detection metrics.  🔸 In contrast, rotation detectors often involve a more complicated loss based on SkewIoU which is unfriendly to gradient-based training.  🔸In this paper, we argue that one effective alternative is to devise an approximate loss that can achieve trend-level alignment with SkewIoU loss instead of the strict value-level identity.  🔸 Specifically, we model the objects as Gaussian distribution and adopt Kalman filter to inherently mimic the mechanism of SkewIoU by its definition, and show its alignment with the SkewIoU at trend-level.  🔸 This is in contrast to recent Gaussian modelling based rotation detectors e.g. GWD, KLD that involves a human-specified distribution distance metric which requires additional hyperparameter tuning.  🔸 The resulting new loss called KFIoU is easier to implement and works better compared with exact SkewIoU, thanks to its full differentiability and ability to handle the non-overlapping cases.  🔸 We further extend our technique to the 3-D case which also suffers from the same issues as 2-D detection. Extensive results on various public datasets (2-D/3-D, aerial/text/face images) with different base detectors show the effectiveness of our approach.  #computervision #artificialintelligence #innovation