3D Instance Segmentation

This task involves detecting and segmenting the object in an 3D scan mesh, which is obtained from the laser scanner point cloud. Submissions must provide a list of 3D instances with their semantic labels and predicted confidences, as well as the mask indicating the vertices belonging to each instance. teaser

Evaluation and Metrics

The set of instance classes is a subset of the 100 semantic classes that includes only the object classes that are countable.

Predicted labels are evaluated per-vertex over the vertices of 5% decimated 3D scan mesh (mesh_aligned_0.05.ply); for 3D approaches that operate on other representations like grids or points, the predicted labels should be mapped onto the mesh vertices.

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.

Evaluation excludes the vertices which are anonymized. The list of these vertices is in mesh_aligned_0.05_mask.txt

Results

Methods MEAN BACKPACK BAG BASKET BED BINDER BLANKET BLINDS BOOK BOOKSHELF BOTTLE BOWL BOX BUCKET CABINET CEILING LAMP CHAIR CLOCK COAT HANGER COMPUTER TOWER CONTAINER CRATE CUP CURTAIN CUSHION CUTTING BOARD DOOR EXHAUST FAN FILE FOLDER HEADPHONES HEATER JACKET JAR KETTLE KEYBOARD KITCHEN CABINET LAPTOP LIGHT SWITCH MARKER MICROWAVE MONITOR MOUSE OFFICE CHAIR PAINTING PAN PAPER BAG PAPER TOWEL PICTURE PILLOW PLANT PLANT POT POSTER POT POWER STRIP PRINTER RACK REFRIGERATOR SHELF SHOE RACK SHOES SINK SLIPPERS SMOKE DETECTOR SOAP DISPENSER SOCKET SOFA SPEAKER SPRAY BOTTLE STAPLER STORAGE CABINET SUITCASE TABLE TABLE LAMP TAP TELEPHONE TISSUE BOX TOILET TOILET BRUSH TOILET PAPER TOWEL TRASH CAN TV WHITEBOARD WHITEBOARD ERASER WINDOW
BFL 0.328 0.473 0.064 0.000 0.387 0.000 0.194 0.684 0.047 0.480 0.213 0.008 0.226 0.121 0.465 0.815 0.843 0.666 0.273 0.533 0.000 0.040 0.475 0.326 0.081 0.002 0.782 0.389 0.000 0.036 0.803 0.431 0.011 0.020 0.785 0.284 0.497 0.020 0.000 0.410 0.859 0.477 0.900 0.019 0.005 0.142 0.001 0.159 0.593 0.664 0.142 0.065 0.000 0.102 0.226 0.060 0.073 0.187 0.000 0.146 0.694 0.156 0.753 0.608 0.182 0.888 0.007 0.001 0.033 0.354 0.176 0.506 0.500 0.298 0.518 0.035 0.828 0.435 0.302 0.193 0.735 0.921 0.950 0.612 0.184
Jiahao Lu, Jiacheng Deng, Tianzhu Zhang. Beyond the Final Layer: Hierarchical Query Fusion Transformer with Agent-Interpolation Initialization for 3D Instance Segmentation. Under Review
SoftGroup 0.297 0.629 0.068 0.000 0.280 0.000 0.532 0.493 0.030 0.447 0.148 0.031 0.133 0.071 0.383 0.736 0.798 0.500 0.000 0.313 0.001 0.013 0.360 0.623 0.021 0.000 0.644 0.070 0.000 0.000 0.756 0.207 0.000 0.166 0.734 0.053 0.364 0.022 0.000 0.816 0.590 0.758 0.940 0.015 0.000 0.002 0.000 0.197 0.265 0.777 0.337 0.000 0.000 0.000 0.104 0.000 0.095 0.152 0.000 0.189 0.823 0.097 0.803 0.666 0.292 0.888 0.001 0.001 0.000 0.454 0.427 0.251 0.283 0.433 0.368 0.119 0.808 0.755 0.130 0.066 0.516 0.813 0.911 0.000 0.177
Thang Vu, Kookhoi Kim, Tung M. Luu, Xuan Thanh Nguyen, Chang D. Yoo. Softgroup for 3d instance segmentation on point clouds. CVPR 2022
Open3DIS - only2D 0.204 0.000 0.285 0.000 0.220 0.116 0.000 0.141 0.000 0.000 0.145 0.133 0.158 0.078 0.036 0.277 0.310 0.763 0.101 0.128 0.000 0.038 0.313 0.103 0.476 0.133 0.446 0.036 0.000 0.600 0.108 0.035 0.010 0.250 0.638 0.033 0.505 0.000 0.000 0.390 0.510 0.325 0.446 0.046 0.000 0.155 0.155 0.000 0.492 0.168 0.072 0.021 0.000 0.206 0.488 0.000 0.344 0.252 0.000 0.034 0.472 0.046 0.117 0.074 0.119 0.364 0.000 0.074 0.381 0.364 0.288 0.112 0.000 0.275 0.368 0.654 0.549 0.107 0.238 0.239 0.480 0.372 0.434 0.277 0.047
Phuc Nguyen, Tuan Duc Ngo, Evangelos Kalogerakis, Chuang Gan, Anh Tran, Cuong Pham, Khoi Nguyen. Open3DIS: Open-vocabulary 3D Instance Segmentation with 2D Mask Guidance. CVPR 2024
HAIS 0.199 0.209 0.000 0.000 0.095 0.000 0.104 0.407 0.029 0.425 0.074 0.000 0.066 0.000 0.312 0.712 0.791 0.333 0.000 0.416 0.000 0.000 0.273 0.430 0.000 0.000 0.514 0.000 0.000 0.000 0.778 0.106 0.000 0.364 0.294 0.090 0.000 0.000 0.000 0.213 0.635 0.548 0.949 0.000 0.000 0.000 0.000 0.000 0.339 0.403 0.375 0.000 0.000 0.000 0.000 0.000 0.000 0.013 0.000 0.170 0.861 0.000 0.882 0.000 0.056 0.666 0.000 0.000 0.000 0.165 0.230 0.242 0.000 0.000 0.217 0.000 0.644 0.000 0.000 0.080 0.376 0.777 0.902 0.000 0.178
Shaoyu Chen, Jiemin Fang, Qian Zhang, Wenyu Liu, Xinggang Wang. Hierarchical aggregation for 3d instance segmentation. ICCV 2021
PointGroup 0.146 0.000 0.000 0.000 0.137 0.000 0.000 0.626 0.012 0.480 0.021 0.000 0.029 0.000 0.222 0.721 0.790 0.000 0.000 0.288 0.000 0.000 0.279 0.334 0.000 0.000 0.460 0.000 0.000 0.000 0.722 0.179 0.000 0.000 0.225 0.033 0.000 0.000 0.000 0.031 0.566 0.490 0.906 0.000 0.000 0.000 0.000 0.000 0.366 0.389 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.203 0.733 0.000 0.000 0.000 0.000 0.535 0.000 0.000 0.000 0.320 0.000 0.196 0.000 0.000 0.000 0.000 0.035 0.000 0.000 0.000 0.494 0.503 0.857 0.000 0.134
Li Jiang, Hengshuang Zhao, Shaoshuai Shi, Shu Liu, Chi-Wing Fu, Jiaya Jia. Pointgroup: Dual-set point grouping for 3d instance segmentation. CVPR 2020

Please refer to the submission instructions before making a submission

Submit results