Design of visual search applications using deep learning

• Main Author: Lim, Joeie Li Xiang
• Other Authors: Yap Kim Hui
• Format: Final Year Project
• Language: English
• Published: 2018
• Subjects: DRNTU::Engineering
• Online Access: http://hdl.handle.net/10356/74779
• Institution: Nanyang Technological University

The advancement of technology has enable electronics manufactures to fit huge computational power into small forms factor at relatively affordable price for general members of public. With that it has led to the next big wave of Smart Nation For example, smart phones are currently an irreplaceable gadget as it is integrated in our everyday life example using it as a reminder for any upcoming event, capturing high quality image without using professional setup etc. Hence, smart gadget could be deployed as a platform for providing search services which could led to big impact in contributing to the building of a Smart Nation. Deep learning has seen huge improvement since earlier time. The objective of this project is to explore the alternative search methodologically through a branch of deep learning: Convolution Neural Network(CNN), enhancing searching experience for its user during the process of image searching. The trained database model that was trained using CNN where it would be able to identify and classify the object based on the images that want loaded into the model. The advantages of this object detection using deep learning are that it is unique and effective as searching the network using images is much more effective rather than describing things in form of words or sentences especially in the application fields such as shopping, browsing, etc. However, the disadvantages of this method are that it may tremendously slow down the process of searching as users will be required to capture and store the image before the application will be able to run its recognition function which may lower the user’s experience. The goal of this project is to train a suitable database for shoes recognition in E commerce application for example shopping of shoes, commercial advantages and etc. Using deep learning on CNN model, which allows the user to obtain the shoes information, as such enhancing their searching experience as it provide an alternative to traditional searching making it more convenient during browsing and doing online purchases. This database would provide a more efficient way of identifying the different types of shoes which helps the user to ease the searching process. Gathering from the result of the project, the database proved to be able to achieve an overwhelm accuracy of 90% and above for searching a specific pair of shoes.