Build a Semi-automatic App for Accurate Crowd Counting by using swift

Finish on or before 22/3. A Semi-automatic App for Accurate Crowd Counting Description: Peaceful and legal rallies serve important functions in a civic society. However, sometimes the figures claimed by different parties are very different. Given that we can video the whole rally at a checkpoint and count manually, this situatiion can certainly be improved. A more accurate figure that is available immediately after the rally and keep being updated is desired. This project aims to develop an accurate people counting app using computer vision and machine learning techniques (e.g. stereo depth sensing and moving object detection available in open source). It counts the number of people in the moving crowd at a checkpoint. In the past year, an app has been developed that enables the user to draw a line, pause the video and annotate the heads, and resume and track the heads. In this way, people can be tracked and counted. In this year, you shall further develop the app to make it a functional app that can be distributed in Apple store. The app will be able to cut the video into fragments and allow users to count fragments of short durations, then sum up the counts for each fragment. It shall also have other functions such as allowing more than one user to count the same fragment. Another aim is to improve the counting accuracy by using more advanced machine learning and artificial intellgence techniques in Apple's CoreML. The knowledge gained for people recognition and counting may also be useful for jobs related to such applications in local vision companies. This is a software project using ipad/iphone. The platform i am using is Xcode. There is also a sample app can be downloaded on App Store, which called People Counter By AI, [login to view URL]