Future Directions and Further Implementations
Because of time limitations, although we were able to use our algorithm to identify cars and parking spots on the street, we were not able to generate enough data to actually train a ML model. If we had more time, we would be able to set some sort of annotating system, to clearly accept/reject what a car leaving or entering looks like in our algorithm, which could eventually allow us to gather a much better data set.
Overall, while our method is a little crude, it is capable of recognizing what it needs to, as well as generate meaningful data. We just needed more time to gather it.
Because of hardware limitations, Faster RCNN was not able to be used. However, if we can successfully utilize either of our algorithms to generate enough data, we need to create a database to store actual data and create machine learning model based on the dataset obtained. Rather than a general bool status of parking availability, the algorithm will be able to obtain the exact number of how many spots are left. This could then be used by a recursive model.
It is also necessary to set up server to push real time data to firebaseDB on a 10 minutes base, data can either be the availability bool value or the exact number of how many spots are left on test sites.
Login with username and password along with sign up services on the mobile application needs to be completed as well after server modification. The application will be added the function that can get parking availability data in a given time on certain site, which is given by user. For example, user can input the estimated arrival time on one site to get a sense of ‘will there be any spots left if I leave 10 min later then expected’ to have a better travel arrangement.
Overall, while our method is a little crude, it is capable of recognizing what it needs to, as well as generate meaningful data. We just needed more time to gather it.
Because of hardware limitations, Faster RCNN was not able to be used. However, if we can successfully utilize either of our algorithms to generate enough data, we need to create a database to store actual data and create machine learning model based on the dataset obtained. Rather than a general bool status of parking availability, the algorithm will be able to obtain the exact number of how many spots are left. This could then be used by a recursive model.
It is also necessary to set up server to push real time data to firebaseDB on a 10 minutes base, data can either be the availability bool value or the exact number of how many spots are left on test sites.
Login with username and password along with sign up services on the mobile application needs to be completed as well after server modification. The application will be added the function that can get parking availability data in a given time on certain site, which is given by user. For example, user can input the estimated arrival time on one site to get a sense of ‘will there be any spots left if I leave 10 min later then expected’ to have a better travel arrangement.