The project will be a part of a larger encompassing project in computer vision. The project will conduct a study to evaluate new and currently used neural network architectures on accuracy, speed, energy efficiency, etc to find better ways of drawing and labeling bounding boxes on a picture of a surface cluttered with trained-on items.
The final goal of the project is to find/implement a better/improved robust neural network for detecting and labeling objects on a cluttered surface. This model will be implemented in future projects that combine its output with a depth map to allow a robot to discern what objects are where and how they are positioned so that it may pick up specific items.