Back to Leaderboard
Parameters
20.0M
FLOPs
60.0G
Input Size
640px
License
Apache-2.0
Architecture
Type
transformer
Backbone
ResNet-18
Neck
HybridEncoder
Head
DETR
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | mAP@50-95 | FPS | Latency | VRAM |
|---|---|---|---|---|
| NVIDIA A100 (TensorRT FP16) | 46.5% | 114.9 | 8.7ms | 914 MB |
| NVIDIA T4 (TensorRT FP16) | 46.5% | 45.6 | 21.9ms | 868 MB |
| CPU (ONNX Runtime) | 46.5% | 5.3 | 187.9ms | 996 MB |
Speed Breakdown (A100 TensorRT)
End-to-end latency breakdown showing preprocessing, inference, and postprocessing times
1.1ms
4.9ms
2.7ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import YOLO
# Load model
model = YOLO.from_pretrained("https://huggingface.co/Libre-YOLO/rtdetr-r18")
# Run inference
results = model.predict("image.jpg")
# Process results
for box in results.boxes:
print(f"Class: {box.cls}, Confidence: {box.conf:.2f}")transformerno-nms
Notes
First real-time DETR, no NMS required
Related Models (rtdetr)