The 2D semantic labeling task involves predicting a per-pixel semantic labeling of an image.

Evaluation and metrics

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



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


Method Infoavg ioubathtubbedbookshelfcabinetchaircountercurtaindeskdoorfloorotherfurniturepicturerefrigeratorshower curtainsinksofatabletoiletwallwindow
sorted bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort bysort by
SSMAcopyleft0.577 10.695 10.716 20.439 40.563 10.314 30.444 10.719 10.551 10.503 10.887 30.346 10.348 20.603 10.353 30.709 10.600 20.457 10.901 10.786 10.599 1
Abhinav Valada, Rohit Mohan, Wolfram Burgard: Self-Supervised Model Adaptation for Multimodal Semantic Segmentation. arXiv
FuseNetpermissive0.521 20.591 30.682 30.220 70.488 30.279 40.344 50.610 30.461 30.475 20.910 10.293 20.447 10.512 30.397 20.618 20.567 40.452 20.734 50.782 20.566 2
Caner Hazirbas, Lingni Ma, Csaba Domokos, Daniel Cremers: FuseNet: Incorporating Depth into Semantic Segmentation via Fusion-based CNN Architecture. ACCV 2016
AdapNet++copyleft0.503 30.613 20.722 10.418 50.358 70.337 20.370 40.479 50.443 40.368 50.907 20.207 50.213 60.464 50.525 10.618 20.657 10.450 30.788 30.721 40.408 6
Abhinav Valada, Rohit Mohan, Wolfram Burgard: Self-Supervised Model Adaptation for Multimodal Semantic Segmentation. arXiv
3DMV (2d proj)0.498 40.481 50.612 40.579 20.456 40.343 10.384 20.623 20.525 20.381 40.845 40.254 40.264 40.557 20.182 50.581 50.598 30.429 40.760 40.661 60.446 5
Angela Dai, Matthias Niessner: 3DMV: Joint 3D-Multi-View Prediction for 3D Semantic Scene Segmentation. ECCV'18
ILC-PSPNet0.475 50.490 40.581 50.289 60.507 20.067 70.379 30.610 30.417 60.435 30.822 60.278 30.267 30.503 40.228 40.616 40.533 50.375 50.820 20.729 30.560 3
Enet (reimpl)0.376 60.264 70.452 70.452 30.365 50.181 50.143 70.456 60.409 70.346 60.769 70.164 60.218 50.359 60.123 70.403 70.381 70.313 70.571 60.685 50.472 4
Re-implementation of Adam Paszke, Abhishek Chaurasia, Sangpil Kim, Eugenio Culurciello: ENet: A Deep Neural Network Architecture for Real-Time Semantic Segmentation.
ScanNet (2d proj)permissive0.330 70.293 60.521 60.657 10.361 60.161 60.250 60.004 70.440 50.183 70.836 50.125 70.060 70.319 70.132 60.417 60.412 60.344 60.541 70.427 70.109 7
Angela Dai, Angel X. Chang, Manolis Savva, Maciej Halber, Thomas Funkhouser, Matthias Nie├čner: ScanNet: Richly-annotated 3D Reconstructions of Indoor Scenes. CVPR'17