All Hardware
NVIDIA A100
NVIDIA A100-PCIE-40GB
VRAM
40 GB
FP32 Performance
19.5 TFLOPS
Power (TDP)
250W
Highest Accuracy
Top 5 models by mAP@50-95
Fastest
Top 5 models by throughput
Results are contributor-reported. Validation checks the submission format and supported configuration; it does not independently reproduce a run. Compare accuracy on the same dataset and latency with the software versions recorded in each run JSON.
All Models
13 models benchmarked on NVIDIA A100 (PyTorch FP32)
| # | Model | Dataset | Credit / source | mAP | FPS | Latency | Params |
|---|---|---|---|---|---|---|---|
| 1 | dfine-x | full (5,000 images) | 59.3% | 15.4 | 64.7ms | 62.6M | |
| 2 | dfine-l | full (5,000 images) | 57.3% | 17.1 | 58.3ms | 31.2M | |
| 3 | dfine-m | full (5,000 images) | 55.1% | 21.7 | 46.1ms | 19.6M | |
| 4 | rtdetr-r101 | full (5,000 images) | 53.9% | 18.6 | 53.6ms | 76.6M | |
| 5 | rtdetr-r50 | full (5,000 images) | 52.7% | 22.7 | 44.1ms | 42.9M | |
| 6 | yolonas-l | full (5,000 images) | 51.2% | 16.6 | 60.4ms | 67.0M | |
| 7 | rtdetr-r50m | full (5,000 images) | 50.8% | 25.1 | 39.8ms | 36.6M | |
| 8 | dfine-s | full (5,000 images) | 50.7% | 24.0 | 41.7ms | 10.3M | |
| 9 | yolonas-m | full (5,000 images) | 50.5% | 17.0 | 58.9ms | 51.2M | |
| 10 | rtdetr-r34 | full (5,000 images) | 48.2% | 27.8 | 36.0ms | 31.4M | |
| 11 | yolonas-s | full (5,000 images) | 46.5% | 16.6 | 60.1ms | 19.1M | |
| 12 | rtdetr-r18 | full (5,000 images) | 45.6% | 27.4 | 36.5ms | 20.2M | |
| 13 | dfine-n | full (5,000 images) | 42.8% | 25.6 | 39.1ms | 3.8M |