Submission: CosyPose-ECCV20-SYNT+REAL-1VIEW/TUD-L

Download submission
Submission name
Submission time (UTC) May 4, 2022, 4:09 p.m.
User yann_labbe
Task Model-based 2D detection of seen objects
Dataset TUD-L
Training model type Default
Training image type Synthetic + real
Description
Evaluation scores
AP:0.826
AP50:1.000
AP75:0.968
AP_large:0.849
AP_medium:0.828
AP_small:0.750
AR1:0.852
AR10:0.852
AR100:0.852
AR_large:0.902
AR_medium:0.844
AR_small:0.750
average_time_per_image:0.041

Method: CosyPose-ECCV20-SYNT+REAL-1VIEW

User yann_labbe
Publication Labbé et al, CosyPose: Consistent multi-view multi-object 6D pose estimation, ECCV 2020
Implementation https://github.com/ylabbe/cosypose
Training image modalities RGB
Test image modalities RGB
Description

The method is the same as CosyPose-ECCV20-PBR-1VIEW but we also add the additionnal real and synthetic images to the training data when an official training split is available: TUD-L, T-LESS and YCB-Video. On other datasets, the results reported are the same as CosyPose-ECCV20-1VIEW-PBR.

The models (detectors, coarse pose estimation, refiner) are pre-trained from the models trained on PBR images only.

Computer specifications CPU: 20-core Intel Xeon 6164 @ 3.2 GHz, GPU: Nvidia V100