Model card
Co-DINO (ViT-L).
Sensetime / Sense-Xopen-source~600M paramsDINO transformer detector with ViT-L backbone and collaborative hybrid assignment training
Co-DINO with ViT-L backbone, largest variant from Co-DETR paper. COCO val2017: 65.9 box AP. LVIS val: 68.0 box AP. Trained with Objects365 pre-training. ICCV 2023. arxiv:2211.12860.
§ 01 · Benchmarks
Every benchmark Co-DINO (ViT-L) has a recorded score for.
| # | Benchmark | Area · Task | Metric | Value | Rank | Date | Source |
|---|---|---|---|---|---|---|---|
| 01 | LVIS v1.0 | Computer Vision · Object Detection | box-ap | 68.0% | #3 | 2022-11-23 | source ↗ |
Rank column shows this model’s position vs all other models scored on the same benchmark + metric (competitors after the slash). #1 in red means current SOTA. Sorted by rank, then newest result.
§ 02 · Strengths by area
Where Co-DINO (ViT-L) actually performs.
§ 03 · Papers
1 paper with results for Co-DINO (ViT-L).
- 2022-11-23· Computer Vision· 1 result
DETRs with Collaborative Hybrid Assignments Training
§ 05 · Sources & freshness
Where these numbers come from.
arxiv
1
result
1 of 1 rows marked verified.