smart-shelf-tracker

YOLO11n Retail Empty Shelf Detector

This repository hosts a custom-trained YOLO11n model, specifically designed for detecting empty shelves in Retails. The dataset used for training was sourced from RoboFlow, titled Retail Empty Shelf Detector.

Model Overview

The model is built on YOLO11n, a lightweight version of the YOLO (You Only Look Once) family of object detectors. It is optimized for real-time object detection, which makes it highly suitable for deployment on edge devices with limited computational power.

• Framework: PyTorch
• Model Type: YOLO11n
• Dataset: RoboFlow - Supermarket Empty Shelf Detector
• Input Size: 640x640 (can be modified)
• Output: Bounding boxes and class labels indicating whether a shelf is empty or not.

Dataset

The dataset used for this model is the Retail Empty Shelf Detector dataset from RoboFlow. It contains annotated images of Retail shelves, both stocked and empty, for the purpose of detecting shelf status.

• Number of Images: 497
• Classes: OOS (Out of Stock)
• Preprocessing: Images are resized to 640x640 for training.

Usage

You can use this model for inference as follows:

from models import YOLO

# Load the model
model = YOLO('<path-to-downloaded-model>')

# Run inference
results = results = model('<path-to-image>')

Improvements

Future improvements planned for this model include:

1. Experiments with Other YOLO11 Models: We will explore other YOLO11 variants (YOLO11s, YOLO11m, etc.) to find the best balance between accuracy and inference speed.
2. Hyperparameter Tuning: Perform extensive hyperparameter optimization (batch size, learning rate, etc.) to enhance the model’s performance.
3. Data Augmentation: Experiment with advanced augmentation techniques such as CutMix, Mosaic, or MixUp to increase the model's generalization capability.

Validation Metrics