Back to Leaderboard

YOLO-NAS-S

yolo-nas

one-stage detector with NAS-Optimized backbone

Parameters

12.2M

FLOPs

32.8G

Input Size

640px

License

Apache-2.0

Architecture

Type

one-stage

Backbone

NAS-Optimized

Neck

QSP

Head

Decoupled

Benchmark Results
Performance on COCO val2017 across different hardware configurations
HardwaremAP@50-95FPSLatencyVRAM
NVIDIA A100 (TensorRT FP16)47.6%144.56.9ms527 MB
NVIDIA T4 (TensorRT FP16)47.4%60.316.6ms566 MB
CPU (ONNX Runtime)47.6%6.8147.0ms557 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.2ms
3.1ms
2.6ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

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

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

# Process results
for box in results.boxes:
    print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")
nas-optimizedefficient
Notes

Neural Architecture Search designed by Deci AI

Related Models (yolo-nas)