Back to Leaderboard

RT-DETR-R50

rtdetr

transformer detector with ResNet-50 backbone

Parameters

42.0M

FLOPs

136.0G

Input Size

640px

License

Apache-2.0

Architecture

Type

transformer

Backbone

ResNet-50

Neck

HybridEncoder

Head

DETR

Benchmark Results
Performance on COCO val2017 across different hardware configurations
HardwaremAP@50-95FPSLatencyVRAM
NVIDIA A100 (TensorRT FP16)53.1%78.012.8ms1861 MB
NVIDIA T4 (TensorRT FP16)53.0%31.631.6ms1875 MB
CPU (ONNX Runtime)53.2%3.6277.2ms1925 MB
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.3ms
8.9ms
2.6ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO

# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/rtdetr-r50")

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

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