In this experiment, reported accuracies were achieved using only first four states. We attempted to discriminate between seven different states of an intersection using still clips at an oblique angle. However, more footage of the last three stages is neccessary to procede.

Nearest-neighbor classifier using Martin distance.

Nearest-neighbor classifier using state KL divergence

SVM classifier using state KL kernel

Nearest-neighbor classifier using image KL divergence

SVM classifier using image KL kernel

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using clips at an oblique angle.

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using clips at a front-ground angle.

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using clips at a side-ground angle.

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using panning clips at an oblique angle.

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using panning clips at a front angle.

In this experiment, we attempt to discriminate between different levels of traffic flow in an intersection using panning clips at a side angle.