The 3D semantic instance prediction task involves detecting and segmenting the object in an 3D scan mesh.

Evaluation and metrics

Our evaluation ranks all methods according to the average precision for each class. We report the mean average precision AP at overlap 0.25 (AP 25%), overlap 0.5 (AP 50%), and over overlaps in the range [0.5:0.95:0.05] (AP). Note that multiple predictions of the same ground truth instance are penalized as false positives.



This table lists the benchmark results for the 3D object detection with limited reconstructions scenario.




Method Infoavg ap 50%bathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
sort bysort bysort bysort bysort bysort bysort bysort bysorted bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
CSC_LR_DET0.364 21.000 10.713 20.468 30.110 10.644 30.078 10.162 30.433 10.182 20.194 30.009 20.333 20.435 10.160 10.715 40.200 40.597 40.123 2
Scratch_LR_DET0.294 40.667 40.672 40.319 40.071 40.552 40.053 30.169 20.360 20.110 40.094 40.007 30.325 30.012 30.083 30.717 30.285 20.764 10.036 4
WS3D_LR_ODpermissive0.370 11.000 10.803 10.702 10.082 20.671 10.028 40.109 40.332 30.189 10.259 10.022 10.351 10.009 40.139 20.738 20.291 10.758 20.176 1
Kangcheng Liu: WS3D: Weakly Supervised 3D Scene Segmentation with Region-Level Boundary Awareness and Instance Discrimination. European Conference on Computer Vision (ECCV), 2022
PointContrast_LR_DET0.338 31.000 10.707 30.507 20.077 30.669 20.075 20.177 10.317 40.155 30.198 20.001 40.277 40.016 20.050 40.834 10.247 30.663 30.111 3