Back to Leaderboard

YOLOX-X

yolox

one-stage detector with CSPDarknet backbone

Parameters

99.1M

FLOPs

281.9G

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)51.1%50.020.0ms4181 MB
NVIDIA T4 (TensorRT FP16)51.0%20.149.8ms4229 MB
CPU (ONNX Runtime)51.2%2.2457.0ms4179 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.3ms
16.1ms
2.7ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

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

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

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