Fruits Detections Using Single Shot MultiBox Detector
Md Liakat Ali, Chris Keller, Michael Huang · 2023
Object detection, specifically, deals with the computer vision task of detecting instances of an object such as humans, animals, or cars in the form of digital images. Single Shot multiBox Detector (SSD) MobileNet model provides confidence values for a set of items. We consider Open Images V6, a supervised dataset consisting of 1.7 million images with 600 categories and images from Fruits 360 from Kaggle for this investigation. Based on the datasets used in this research, the model was able to identify selected fruits. With sufficient lighting conditions, the model performed with a confidence value between 75% and 99%. While this experiment produces favorable results under ideal lighting conditions, under suboptimal lighting conditions the model's performance drops. This creates complications for future implementations when creating a model for detecting pedestrians in self-driving cars due to a variety of lighting conditions experienced outdoors.