This project implements a YOLO-style object detection model
from scratch for detecting fruits in images.

It is inspired by automatic checkout systems such as
Amazon “Just Walk Out”, where real-time object detection
is used to identify and track products.

The main goal is to understand how YOLO works internally,
including CNN-based feature extraction, grid-based prediction,
IoU-based responsibility assignment, and loss optimization.

The model takes an RGB image as input, divides it into
a fixed grid, and predicts bounding boxes, confidence scores,
and class probabilities in a single forward pass.

This single-pass design enables fast, end-to-end detection
and is suitable for real-time applications.

The project follows the COCO 2017 object detection format
and uses a filtered subset of fruit categories
(apple, banana, orange). The dataset is not included in this repository.
COCO 2017 dataset:
https://www.kaggle.com/datasets/awsaf49/coco-2017-dataset