Background: Plastibot revolves around the development of a machine learning model designed to classify various types of plastic using photos captured by a hyperspectral camera. The project’s initial phase involves collecting data by cutting up plastic samples and subjecting them to the hyperspectral camera. The primary objective currently is to identify a ‘pivot,’ which could lead the project in one of two directions: refining machine learning methods to exclusively sort plastic or exploring alternatives, such as utilizing regular phone cameras and employing image processing techniques for plastic sorting.
Project Goal: Plastibot is a team focused on revolutionizing plastic classification by enhancing the accuracy of hyperspectral imaging and Convolutional Neural Networks (CNNs). Our current CNN is capable of leveraging the chemical information hidden in hyperspectral images to classify the six main types of plastics with enhanced accuracy. By identifying which spectral bands correlate with specific plastic types, we aim to develop a cost-efficient solution that can be deployed in real-world environments, such as landfills. This approach will allow us to replace expensive hyperspectral cameras with specialized cameras targeting key spectral bands.
Focus Areas: Machine learning, hyperspectral imaging, image classification, data collection and analysis, algorithm development, image processing, neural network operation
