Waste Object Detection

Project overview

Waste images were annotated for a computer vision dataset focused on detecting and classifying different types of waste material. The work involved creating accurate bounding box annotations around waste objects inside bins and in the surrounding environment, producing a high-quality dataset for AI-based object detection models.

Work performed

• Reviewed and organised the waste image dataset before annotation.
• Created precise bounding boxes around individual waste objects.
• Multiple waste categories were annotated while class consistency was maintained.
• Handled images containing overlapping and partially visible waste items.
• Quality checks identified and corrected annotation mistakes.
• Ensured accurate object localisation for object detection model training.

What made it hard

• Annotating overlapping waste objects inside crowded bins.
• Detecting small objects such as cans, bottles and disposable cups.
• Partially visible and occluded items required careful judgement.
• Placing accurate bounding boxes around irregularly shaped objects.
• Differentiating between visually similar waste categories.

Result

A high-quality waste detection dataset covering fourteen waste object classes was completed and delivered with accurate bounding box annotations. The dataset was prepared for AI-based object detection, waste classification and automated recycling applications.

Final deliverable

An accurately annotated waste detection dataset was delivered with bounding boxes across the different waste categories, holding high annotation quality and consistency. The final dataset was ready for training computer vision models for automated waste detection and recycling systems.