The HandCam system has two unique properties as compared to egocentric systems: it avoids the need to detect hands, and it more consistently observes the activities of hands. By taking advantage of these properties, we collected a HandCam-Object Dataset. In our experiment, accuracy of HandCam is about 10% higher than accuracy of egocentric camera.
Cheng-Sheng Chan, Ting-An Chien, Tz-Ying Wu, Min Sun
Asian Conference on Computer Vision(ACCV), 2016
Augmented Reality demo
This project is a simple AR demo by using OpenCV. I draw a simple model(the little box man) and display it on the scene capture by the webcam. I get the camera extrinsic parameter by detecting the AR tag as reference, ,then transform the 3D model to 2D and display it on the screen.
GoodoBaby: An intelligent baby care system
This project is a prototype of an IOT system. I and my partner build the whole system. On the backend we set up a LAMP server, which handles all the services from our clients (sensing devices and phone). For the physical devices, we use Ameba RTL8195 and combine it with several sensors (ultrasonic, accelerometer, etc). And the last, we write an android APP for users to interact with the whole system.
Part-time Intern in PROSCEND, Hsinchu, Taiwan
July. 2015 - June. 2016