Ashish Patel 🇮🇳’s Post

𝗗𝗮𝘆-𝟯𝟳𝟮 𝗖𝗼𝗺𝗽𝘂𝘁𝗲𝗿 𝗩𝗶𝘀𝗶𝗼𝗻 𝗟𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗦𝗔𝗦𝗔: 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰𝘀-𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗦𝗲𝘁 𝗔𝗯𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗣𝗼𝗶𝗻𝘁-𝗯𝗮𝘀𝗲𝗱 𝟯𝗗 𝗢𝗯𝗷𝗲𝗰𝘁 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 𝗯𝘆 𝗧𝗵𝗲 𝗨𝗻𝗶𝘃𝗲𝗿𝘀𝗶𝘁𝘆 𝗼𝗳 𝗦𝘆𝗱𝗻𝗲𝘆 Follow me for a similar post: Ashish Patel  ------------------------------------------------------------------- 𝗜𝗻𝘁𝗲𝗿𝗲𝘀𝘁𝗶𝗻𝗴 𝗙𝗮𝗰𝘁𝘀 : 🔸 Paper: 𝗦𝗔𝗦𝗔: 𝗦𝗲𝗺𝗮𝗻𝘁𝗶𝗰𝘀-𝗔𝘂𝗴𝗺𝗲𝗻𝘁𝗲𝗱 𝗦𝗲𝘁 𝗔𝗯𝘀𝘁𝗿𝗮𝗰𝘁𝗶𝗼𝗻 𝗳𝗼𝗿 𝗣𝗼𝗶𝗻𝘁-𝗯𝗮𝘀𝗲𝗱 𝟯𝗗 𝗢𝗯𝗷𝗲𝗰𝘁 𝗗𝗲𝘁𝗲𝗰𝘁𝗶𝗼𝗻 🔸 This paper is published arxiv2021. 🔸 Semantics-Augmented Set Abstraction (SASA) for point-based 3D detection. Our main concept is to incorporate semantic information into the Pointnet SA stage for guiding the point-based backbone to better model potential objects. Experimental results on the KITTI and nuScenes datasets indicate that our strategy can help access a higher point recall during the point downsampling stage so as to obtain a better detection outcome for manifold point-based detectors. ------------------------------------------------------------------- 𝗜𝗠𝗣𝗢𝗥𝗧𝗔𝗡𝗖𝗘 🔸 Although point-based networks are demonstrated to be accurate for 3D point cloud modeling, they are still falling behind their voxel-based competitors in 3D detection.  🔸 We observe that the prevailing set abstraction design for down-sampling points may maintain too much unimportant background information that can affect feature learning for detecting objects.  🔸 To tackle this issue, we propose a novel set abstraction method named Semantics-Augmented Set Abstraction (SASA).  🔸 Technically, we first add a binary segmentation module as the side output to help identify foreground points.  🔸 Based on the estimated point-wise foreground scores, we then propose a semantics-guided point sampling algorithm to help retain more important foreground points during down-sampling.  🔸 In practice, SASA shows to be effective in identifying valuable points related to foreground objects and improving feature learning for point-based 3D detection. Additionally, it is an easy-to-plug-in module and able to boost various point-based detectors, including single-stage and two-stage ones.  🔸 Extensive experiments on the popular KITTI and nuScenes datasets validate the superiority of SASA, lifting point-based detection models to reach comparable performance to state-of-the-art voxel-based methods. #computervision #artificialintelligence #innovation