The Scene type classification task involves classifying a scan into 13 scene types.

Evaluation and metrics

Our evaluation ranks all methods according to recall (TP/(TP+FN)) as well as the PASCAL VOC intersection-over-union metric (IoU = TP/(TP+FP+FN)), where TP, FP, and FN are the numbers of true positive, false positive, and false negative predictions, respectively.



This table lists the benchmark results for the scene type classification scenario.




Method Infoavg recallapartmentbathroombedroom / hotelbookstore / libraryconference roomcopy/mail roomhallwaykitchenlaundry roomliving room / loungemiscofficestorage / basement / garage
sort bysort bysort bysort bysort bysort bysort bysorted bysort bysort bysort bysort bysort bysort by
3DASPP-SCE0.691 30.500 10.938 30.824 41.000 11.000 10.500 31.000 10.857 30.500 20.556 40.000 20.812 30.500 2
LAST-PCL-type0.780 10.250 31.000 11.000 11.000 11.000 11.000 10.500 21.000 10.500 20.889 10.000 21.000 11.000 1
Yanmin Wu, Qiankun Gao, Renrui Zhang, and Jian Zhang: Language-Assisted 3D Scene Understanding. arxiv23.12
multi-taskpermissive0.700 20.500 11.000 10.882 30.500 31.000 11.000 10.500 21.000 11.000 10.778 20.000 20.938 20.000 3
Shengyu Huang, Mikhail Usvyatsov, Konrad Schindler: Indoor Scene Recognition in 3D. IROS 2020
SE-ResNeXt-SSMA0.498 40.000 50.812 40.941 20.500 30.500 40.500 30.500 20.429 50.500 20.667 30.500 10.625 40.000 3
Abhinav Valada, Rohit Mohan, Wolfram Burgard: Self-Supervised Model Adaptation for Multimodal Semantic Segmentation. arXiv
resnet50_scannet0.353 50.250 30.812 40.529 50.500 30.500 40.000 50.500 20.571 40.000 50.556 40.000 20.375 50.000 3