Back to Leaderboard

YOLOX-Nano

yolox

one-stage detector with CSPDarknet backbone

Parameters

0.9M

FLOPs

1.1G

Input Size

640px

License

Apache-2.0

Architecture

Type

one-stage

Backbone

CSPDarknet

Neck

PAFPN

Head

Decoupled

Benchmark Results
Performance on COCO val2017 across different hardware configurations
HardwaremAP@50-95FPSLatencyVRAM
NVIDIA A100 (TensorRT FP16)25.8%219.24.6ms39 MB
NVIDIA T4 (TensorRT FP16)25.9%89.711.1ms66 MB
CPU (ONNX Runtime)25.9%11.388.4ms53 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.3ms
0.8ms
2.5ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

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

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

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