Parameters62.0M
GFLOPs202.0
Input Size640px
Best mAP60.0%
LicenseApache-2.0
Architecture
Type
transformer
Backbone
HGNetv2
Neck
HybridEncoder
Head
DETR
Benchmark Results
Performance on COCO val2017 across different hardware configurations
| Hardware | Runtime | mAP@50-95 | FPS | Latency | VRAM |
|---|---|---|---|---|---|
| NVIDIA Jetson Orin Nano Super 8GB | ONNX Runtime FP32 | 60.0% | 2.9 | 342.2ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 60.0% | 2.8 | 351.8ms | 399 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 60.0% | 13.8 | 72.7ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 60.0% | 7.8 | 128.3ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 60.0% | 38.3 | 26.1ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 60.0% | 16.6 | 60.1ms | 404 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 60.0% | 71.0 | 14.1ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 60.0% | 58.3 | 17.1ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
8.8ms
339.8ms
3.2ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibreRTDETRv4x.pt")
# Run inference
result = model("image.jpg", conf=0.25, iou=0.45)
# Process results
print(f"Found {len(result)} objects")
print(result.boxes.xyxy) # bounding boxes (N, 4)
print(result.boxes.conf) # confidence scores (N,)
print(result.boxes.cls) # class IDs (N,)detrnms-freePaper: 57% mAP