The 2D semantic instance prediction task involves detecting and segmenting the object in an image.

Evaluation and metrics

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



This table lists the benchmark results for the 2D semantic instance scenario.




Method Infoavg apbathtubbedbookshelfcabinetchaircountercurtaindeskdoorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwindow
sorted bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
UniDet_RVC0.205 10.381 10.323 10.037 10.226 10.177 10.063 10.277 10.120 10.067 10.131 10.074 20.317 10.080 10.235 10.289 10.141 10.678 10.080 1
MaskRCNN_ScanNetpermissive0.119 20.129 20.212 20.002 20.112 20.148 20.014 20.205 20.044 20.066 20.078 20.095 10.142 20.030 20.128 20.139 20.080 20.459 20.057 2
Re-implementation of Kaiming He, Georgia Gkioxari, Piotr Dollár, Ross Girshick: Mask R-CNN. ICCV'17