Suppose I have different videos of the same car sometimes moving slow, sometimes moving fast, say, at 50Kmph as slow and 60Kmph as fast. (Assume the background is a green screen and the car doesn't cross the camera frame ever, the clip ends much before that)
For example, the following are the sample frames from a video.
I need to classify the videos into slow or fast.
One approach is to feed the video directly to 3D-CNN as a 4-D tensor and train against the output as fast or slow.
I was thinking if we can consider the size of the object in progressive frames as time-series data. Then if we apply time-series classification, could it be a good idea?