The proposed dataset contains a collection of videos acquired using a stationary camera (640x480 pixels). In these sequences we can find mostly pedestrians walking in one of the following directions:
to the left, to the right and to the front. It may be more than one pedestrian in a sequence and in some sequences pedestrians may walk in different directions.
This dataset was classified manually, creating a ground truth split for training and another one for testing. The dataset comprised 7416 images for training and 1752 for validation. Images were classified according to three different categories, right, left and front: 2907 images were assigned to 'Right', 3099 images to 'Left' and 1410 to 'Front'.
Below we can find some example clips that show the different scenarios where the dataset was recorded.
Left
Right
Multiple directions
Zebra crossing
Download
We are preparing data in a standard format so other researchers can evaluate their methods using the proposed dataset.
It will be soon available on this website.
Robotics & Tridimensional Vision Research Group. University of Alicante