Back to Leaderboard

YOLOv9-M

yolov9

one-stage detector with GELAN backbone

Parameters

20.1M

FLOPs

76.8G

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
HardwaremAP@50-95FPSLatencyVRAM
NVIDIA A100 (TensorRT FP16)51.4%117.78.5ms923 MB
NVIDIA T4 (TensorRT FP16)51.4%42.623.5ms890 MB
CPU (ONNX Runtime)51.4%5.4185.4ms879 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.1ms
4.9ms
2.5ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/yolov9m")

# Run inference
results = model.predict("image.jpg")

# Process results
for box in results.boxes:
    print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")
programmable-gradientbalanced
Related Models (yolov9)