Parameters9.0M
GFLOPs13.5
Input Size640px
Best mAP44.3%
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 | 44.3% | 17.1 | 58.6ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 44.3% | 16.9 | 59.2ms | 87 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 44.2% | 34.9 | 28.6ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 44.3% | 26.1 | 38.3ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 44.3% | 60.6 | 16.5ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 44.3% | 50.0 | 20.0ms | 87 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 43.4% | 70.9 | 14.1ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 44.3% | 74.5 | 13.4ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
8.7ms
41.5ms
9.0ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibreYOLOXs.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: 40.5% mAP