Open-vocabulary Object Detection on COCO
42.1AP (Novel)Grounding DINO-T
Evaluation Results
| Method | Links | |
|---|---|---|
| Grounding DINO-TBackbone=Swin-T, Text Enc.=BERT-Base, Latency (ms)=∼280, FPS=3.52026.04 | 42.1 | |
| DeCo-DETRBackbone=ResNet-50, Latency (ms)=135, FPS=7.42026.04 | 41.3 | |
| CAKEBackbone=ResNet-50, Latency (ms)=145, FPS=6.92026.04 | 38.2 | |
| VL-PLMBackbone=ResNet-50, Text Enc.=RoBERTa, Latency (ms)=∼210, FPS=4.72026.04 | 32.3 | |
| DetProBackbone=ResNet-50, Latency (ms)=250, FPS=4.02026.04 | 29.4 |