| Submission name |
X0 (2ddefaultdet)
|
| Submission time (UTC) |
March 22, 2026, 11:42 a.m.
|
| User |
francescamariagreco
|
| Task |
Model-free 6D detection of unseen objects |
| Dataset |
HOPEv2 |
| Description |
Training data: None; Test image modalities(HOPEv2): RGB
Model-free: using dynamic and static onboarding sequences to build 3DGS per object models through COLMAP, generating pose hypotheses with PnP and SuperRANSAC, using DINOV3 and SAM based segmentation, sparse activation for the inference with a threshold based on temporal sequence model to aggregate per-frame features and inclusion of a density-based confidence module based on GPT-2. Computer specifications: AMDmi300A, AMDmi300X.
|
| Evaluation scores |
| AP: | 0.473 |
| AP_25: | 0.385 |
| AP_25_mm: | 0.139 |
| AP_MSPD: | 0.560 |
| AP_MSSD: | 0.385 |
| AP_MSSD_mm: | 0.139 |
| average_time_per_image: | 40.980 |
|