NorgesGruppen Data Challenge - Model Weights
YOLO ensemble weights for grocery shelf product detection (356 categories).
Model Details
| Model | Architecture | Resolution | Size | Format |
|---|---|---|---|---|
| model1.onnx | YOLOv8m | 640px | 94 MB | ONNX |
| model2.onnx | YOLOv8x | 1280px | 249 MB | ONNX |
Performance
- Score: 0.9006 (mAP@0.5)
- Competition: NM i AI 2026 - NorgesGruppen Data Challenge
Usage
import onnxruntime as ort
from PIL import Image
import numpy as np
# Load model
session = ort.InferenceSession("model1.onnx", providers=["CUDAExecutionProvider"])
# Preprocess image (letterbox to 640x640)
img = Image.open("shelf.jpg").convert("RGB")
# ... preprocessing code ...
# Run inference
output = session.run(None, {"images": img_array})[0]
Source Code
Full training and inference code: GitHub - ngd-challenge
License
MIT License