Back to Leaderboard
YOLO11-M
yolo11Coming Soonone-stage detector with CSPDarknet backbone
Parameters20.1M
GFLOPs68.0
Input Size640px
Best mAP50.6%
Licensenon-permissive
Architecture
Type
one-stage
Backbone
CSPDarknet
Neck
C3k2-PAFPN
Head
Decoupled
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | Runtime | Dataset | Credit / source | mAP@50-95 | FPS | Latency | GPU allocation |
|---|---|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | mini500 (500 images) | 50.6% | 1.5 | 686.2ms | - | |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | mini500 (500 images) | 50.6% | 11.1 | 90.3ms | 190 MB | |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | mini500 (500 images) | 50.6% | 25.8 | 38.8ms | 14 MB | |
| Raspberry Pi 5 | ncnn FP32 | mini500 (500 images) | 50.6% | 1.9 | 532.7ms | - | |
| Raspberry Pi 5 | ONNX Runtime FP32 | mini500 (500 images) | 50.6% | 1.0 | 994.5ms | - | |
| Raspberry Pi 5 | PyTorch FP32 | mini500 (500 images) | 50.6% | 0.9 | 1104.3ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
6.3ms
664.6ms
9.3ms
Preprocess
Inference
Postprocess (NMS)
anchor-freenmsPaper: 51.5% mAP