YOLO26 Medium — Retail Product Detector
Single-class product detector trained on the Retail Product Dataset.
All original classes were collapsed into one class: product.
Metrics (val set)
| Metric | Value |
|---|---|
| mAP50 | 0.7876 |
| mAP50-95 | 0.5128 |
| Precision | 0.8496 |
| Recall | 0.8565 |
Usage
from ultralytics import YOLO
from huggingface_hub import hf_hub_download
weights = hf_hub_download(repo_id="prince4332/yolov26-product-detection", filename="best.pt")
model = YOLO(weights)
results = model.predict("your_image.jpg", conf=0.25)
Training Config
- Base model: yolo26m (COCO pretrained)
- Epochs: 80 (early stopping patience=20)
- Image size: 640
- Optimizer: auto (MuSGD)
- AMP: True (FP16)
- Augmentation: mosaic, mixup, HSV, perspective, flips
- Downloads last month
- 257