Most of us have our lives thick in the hubbub of Atlanta city life. But with such a lively city comes even livelier traffic. How to alleviate this problem? Well the first step to solving any problem is to first watch and analyze.
What it does
User can input a date and time and then the street camera and real-time video footage will pull up from the SmartCity API and allow you to see the current traffic on certain streets and times of the day. With Google's object detection API, we are somewhat able to track what different objects are and classify them, such as cars.
How we built it
We used Tensorflow and the GE CityIQ API to retrieve the video data. We had to manipulate the time in Epoch to retrieve certain dates and streets.
Challenges we ran into
The data isn't actually real-time, it is simulated data so the times and streets weren't always 100% accurate, limiting the analysis we could do on the traffic. In fact, some of the video files we received were movie trailers, and audio retrieved included the hit song "Despacito". It was also very difficult merging our two codes together, the Python user interface and the pulling information from the API.
Accomplishments that we're proud of
Finally understanding Postman, especially now that many companies are using it to make and test APIs. Figuring out how to extract the videos we needed and using them was also an immense obstacle we hurdled over. We accomplished using object detection on moving objects (cars, etc) in the video footage with Tensorflow. We learned how to use Postman environment and extract data from APIs. We also learned using Tkinter to program a basic Python user interface, and how to access files through the computer through our code.
What's next for Trafflickr
With the Google object detection API, we want to quantify the data and be able to detect abnormalities and find the risk factor of accidents. So far we are only able to analyze and view video footage. Next, we want to be able to access and detect abnormalities in audio so that it will eventually be able to detect gunshots, screams, accident noises, etc and alert the police. With the video footage we also want to soon be able to use facial recognition analysis on pedestrians to see if they match wanted criminal mugshots or kidnapped children.