Apple Detection

Project overview

Images of apple trees were annotated for a computer vision dataset focused on detecting apples in natural outdoor environments. The work involved identifying individual apples on tree branches and creating accurate bounding box annotations around each visible fruit.

The annotated dataset was prepared as training data for an AI model designed to detect apples automatically in images.

Work performed

• Reviewed and organised the image dataset before annotation began.
• Identified individual apples visible on trees and branches.
• Created accurate bounding boxes around each apple.
• Maintained consistent labelling using the Apple class.
• Partially visible apples were annotated wherever applicable.
• Quality checks were carried out to find and correct inaccurate or missed annotations.
• Annotations were confirmed as suitable for object detection model training.

What made it hard

• Detecting small apples among dense leaves and branches.
• Overlapping and partially occluded apples had to be handled individually.
• Apples appeared at widely different distances and viewing angles.
• Keeping bounding box boundaries tight around individual apples.
• Similarly coloured objects and background elements were easily mistaken for fruit.

Result

A high-quality apple detection dataset was completed with bounding box annotations for individual apples. The dataset was prepared for training an object detection model capable of identifying apples automatically in tree images.

Final deliverable

Accurate bounding box annotations for apples were delivered across the tree images, with consistent labelling and annotation quality maintained throughout the dataset. The final output met the project requirements and was ready for AI model training and development.