litter-detection-tensorflow
Litter-detection-tensorflow is an object detection algorithm built by team fixIT for a CIS490 Project Management final project. It uses the TensorFlow API, Jupyter Notebook, and an AWS Ubuntu server to identify and localize litter in Google Streets images. The system leverages TensorFlow's pre-trained COCO model as a foundation and is further trained on a custom dataset of 10,000 Google Streets images labeled with labelImg software by student volunteers provided by Keep America Beautiful. Key features include pattern recognition to detect objects, object detection that boxes identified items with confidence percentages, classification of litter among detected objects, and an image ranking system from 1 to 4 indicating litter severity. A custom script automates image collection from Google Streets. The goal is to help Keep America Better understand which types of litter occur most in different zip codes, enabling more targeted and efficient community cleanup efforts. The project follows a structured developmen