Count How Many Out There
Knowing the number of objects in an image provides a count on how many objects or the size of the crowd. This information is important to traffic control at intersections that detects the number of vehicles per lane or home security system that detects the number of intruders. There are commercial systems that output the existence of a certain object, but customization is needed to output the amount, i.e., how many. Some coding is required for this project along with constructing a data capture system that includes a capture and a computation unit. A casing box is needed as the system will be tested in indoor and outdoor environments.
Objectives
A Raspberry Pi unit with camera image capture capability and a light Machine Learning image classification tool should be first constructed and tested. The images should be captured using sensor activated mechanisms. Later, the Machine Learning software should be updated to have object counting capability. The system should be tested in a lab environment where the objects are toys, books and furniture, and in an outdoor environment where the objects are plants, animals and vehicles. No humans should be included in the images. The size of the physical system should be small, e.g., palm size.
Motivations
Camera images and videos contain information on what is out there and how many things are there. Counting takes extra computation and the accuracy can suffer due to blurriness and low light. Camera detection systems that provide the counting functionality for outdoor environments are not readily available. This work can better inform camera detection and surveillance system users with an estimate on the type and the amount of the object.
Qualifications
Minimum Qualifications:
- Raspberry Pi (1 month +)
- OpenCV_object detection (2 weeks + )
- Python (2 month +)
Experiences with constructing sensor systems and using Machine Learning tools such as:
- ResNet
- YOLO
- RetinaNet
Details
Project Partner:
Yue Zhang
NDA/IPA:No Agreement Required
Number Groups:2
Project Status:Accepting Applicants
Keywords:Computer VisionResearchHardwareConsultancyMachine Learning MLArtificial Intelligence AI
