Parameters0.9M
GFLOPs1.3
Input Size640px
Best mAP28.8%
LicenseApache-2.0
Architecture
Type
one-stage
Backbone
CSPDarknet
Neck
PAFPN
Head
Decoupled
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 | 28.8% | 34.4 | 29.1ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 28.8% | 18.4 | 54.5ms | 17 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 28.8% | 51.0 | 19.6ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 28.8% | 47.4 | 21.1ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 28.8% | 63.2 | 15.8ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 28.8% | 53.6 | 18.7ms | 17 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 28.7% | 100.0 | 10.0ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 28.6% | 86.7 | 11.5ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
7.6ms
38.0ms
8.9ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibreYOLOXn.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,)anchor-freenmsPaper: 25.8% mAP