Parameters1.0M
GFLOPs0.7
Input Size320px
Best mAP30.4%
LicenseApache-2.0
Architecture
Type
one-stage
Backbone
ESNet
Neck
CSP-PAN
Head
GFL
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 | 29.6% | 16.4 | 61.1ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | PyTorch FP32 | 30.4% | 13.8 | 72.2ms | 12 MB |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP16 | 30.4% | 39.6 | 25.3ms | - |
| NVIDIA Jetson Orin Nano Super 8GB | TensorRT FP32 | 30.4% | 37.9 | 26.4ms | - |
| NVIDIA RTX 5070 Ti | ONNX Runtime FP32 | 30.4% | 55.8 | 17.9ms | - |
| NVIDIA RTX 5070 Ti | PyTorch FP32 | 30.4% | 50.0 | 20.0ms | 22 MB |
| NVIDIA RTX 5070 Ti | TensorRT FP16 | 30.4% | 78.6 | 12.7ms | - |
| NVIDIA RTX 5070 Ti | TensorRT FP32 | 30.3% | 74.0 | 13.5ms | - |
Speed Breakdown(NVIDIA Jetson Orin Nano Super 8GB)
4.3ms
47.3ms
20.7ms
Preprocess
Inference
Postprocess (NMS)
Usage with LibreYOLO
from libreyolo import LibreYOLO
# Load model (auto-downloads from HuggingFace if not found locally)
model = LibreYOLO("LibrePICODETs.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: 27.1% mAP