Vamsi Charan Adari
Download Project (248 KB)
The purpose of this project is to compare different deep learning frameworks used to detect, recognizeobjects and images using mobile devices. In particular, mobile recognizing frameworks recognize objectsbased on features extracted, color patterns and object segments. When a similar object from traineddataset is identified matching to the recognized object, framework checks the accuracy rate among allthe other related objects and translated it as a command and shows the name of approximate detectedobject. First, the camera of the mobile device captures image of the object and send the image toframework to extract features, segment the object based on shape, color and size. Second, the relatedapproximate object from trained dataset is identified and sent as a return command to the user mobile.Third, the approximate name of the object is displayed on the user’s mobile phone helping the user torecognize. To demonstrate the accuracy rate and functioning of the mobile frameworks, we developedthree mobile applications to demonstrate the effectiveness of the new of mobile interaction. All thethree applications take image as an input and shows an approximate output on the mobile to the user.This comparison in mobile frameworks will facilitate the better usability of different mobile recognitionframeworks in mobile devices to recognize better and unknown objects moving forward.
Van Tam Nguyen
Primary Advisor's Department
Stander Symposium project, College of Arts and Sciences
United Nations Sustainable Development Goals
Industry, Innovation, and Infrastructure
"Comparison of Machine Learning Frameworks for Mobile Devices" (2020). Stander Symposium Projects. 1790.