| Summary: | In this era of advanced technology, mobile devices become one of the most fascinating technology that we are using in our daily life. With a mobile device, people can take photo, record video, play games, check weather condition, surf internet for useful information as long as it is connected to the network and use many other mobile applications for the benefit of their life. Hence, mobile devices as well as their mobile applications are gaining popularity and usage of it are growing. In parallel to the growth of mobile devices and its applications, the number of people going outside for shopping are also increasing in order to supplement their daily needs as well as improve their lifestyle. They may go for shopping to buy foods, drinks, health and beauty products, etc. When they go for shopping and buy a product, they may eager to know if they are buying the right kind of product with good quality and fair price. It will be a great help if they can take the photo of their chosen product with their smart phone camera and there is a mobile application which can identify the product from the image taken as well as return some useful information like company profile, quality of the product, recommendation from the professions as well as other users and lastly the price comparison from different retail store.
With the aim to help those people who want to know more about their chosen product, we will study the various image recognition techniques and build the large image database containing different rigid commercial products under the categories of foods, drinks, health and beauty products, etc. to test these techniques. The testing will be done under 5 different test conditions by considering the possible scenario whereby user may take the image of the products with their smart phone camera when they are going outside for shopping. Moreover, same test condition will run with and without Geometric verification (GV) to evaluate and compare the impact to recognition accuracy and run time. Furthermore, 10,000 distractor images will also add into the reference image database to test the performance of the software. In the end of the project, we will provide the suggestion on most suitable condition and method to capture the rigid commercial products in order to get the best possible result. Lastly, we will further access and explore the ways to improve the mobile visual search system for future development.
|
|---|