Parameters54.2M
GFLOPs78.0
Input Size640px
Best mAP55.5%
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 | 55.5% | 6.2 | 161.9ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 55.4% | 5.8 | 171.7ms | 306 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 55.5% | 21.6 | 46.3ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 55.4% | 11.4 | 87.5ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 55.4% | 50.6 | 19.8ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 55.4% | 42.8 | 23.3ms | 307 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 55.4% | 79.3 | 12.6ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 55.4% | 61.8 | 16.2ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
8.8ms
154.4ms
8.5ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibreYOLOXl.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: 49.7% mAP