Back to Leaderboard
Parameters
7.2M
FLOPs
26.7G
Input Size
640px
License
MIT
Architecture
Type
one-stage
Backbone
GELAN
Neck
PGI
Head
Decoupled
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | mAP@50-95 | FPS | Latency | VRAM |
|---|---|---|---|---|
| NVIDIA A100 (TensorRT FP16) | 46.7% | 164.3 | 6.1ms | 343 MB |
| NVIDIA T4 (TensorRT FP16) | 46.9% | 65.7 | 15.2ms | 461 MB |
| CPU (ONNX Runtime) | 46.7% | 7.6 | 131.9ms | 355 MB |
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.2ms
2.5ms
2.3ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO
# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/yolov9s")
# Run inference
results = model.predict("image.jpg")
# Process results
for box in results.boxes:
print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")programmable-gradientefficient