Using a Raspberry Pi and tensorflow lite. We used a model to detect and graph a heatmap of when traffic was highest going in and out of TMU dorm area.
Model taken from https://github.com/anis-13/Cars-detection-on-Tflite/blob/master/model_float.tflite and code modified.
Raw results are found in results.txt where you can see the hours of operations and how many frames were captured of cars passing through. Program running at ~6 FPS so if a car passes by up or down the hill it would be about 3-5 seconds of detection. By dividing the frames captured by these low and high bounds we can get a range of how many vehicles passed by. (each car = 18 - 30 frames). There are some issues with the way we capture parked cars. If a car was parked in front of the camera for an extended amount of time, that may skew the distribution.
We expected to see some sort of normal outcome/distribution. What we can see from the graph is that peak times are 10am-12pm and 4pm-6pm which follows a reverse bell curve look of normality.