IoT Based Object Detection and Identification with Open CV using Web CAM

The efficiency and precision of object detection methods have been substantially enhanced by the latest developments in deep learning. This has allowed for the execution of very precise object detection at lightning speed on state-of-the-art desktop PCs. There has been a recent uptick in research towards creating embedded-device-optimal deep neural network topologies that are both quicker and smaller. A popular embedded computer board, the Raspberry Pi 3 is investigated in this thesis as to whether it is suitable for conducting object detection. In two blind tests, two cutting-edge object recognition models, YOLO and SSD, are put to the test in terms of accuracy and speed. The results demonstrate that YOLO is somewhat more accurate and faster than the SSD model, however none of these models is practical for use in applications requiring high speed due to the present generation of Raspberry Pi's limited processing capability.