Parameters5.1M
GFLOPs7.7
Input Size640px
Best mAP35.5%
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 | 35.5% | 31.4 | 31.8ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 35.5% | 20.3 | 49.2ms | 37 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 35.5% | 51.6 | 19.4ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 35.5% | 43.6 | 23.0ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 35.5% | 76.2 | 13.1ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 35.5% | 61.6 | 16.2ms | 37 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 35.4% | 97.7 | 10.2ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 33.9% | 87.2 | 11.5ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
7.7ms
31.3ms
10.2ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibreYOLOXt.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: 32.8% mAP