WALT

Watch And Learn 2D Amodal Representation using Time-lapse Imagery


Watch And learn Time-Lapse(WALT) Dataset

We introduce a new dataset, Watch and Learn Time-lapse (WALT), consisting of 12 (4K and 1080p) cameras capturing urban environments over a year. [DOWNLOAD]

Update: New images from a DOT "Better Boulevard Analytics" project are now part of the WALT dataset. The new images are from six cameras in a suburban setting. Access them here.

Camera1

Camera2

Camera3

Camera4

Camera5

Camera6

Camera7

Camera8

Suburb Image Dataset

Supplemental images from six additional cameras located in a suburb. All of the cameras are mounted approximately 10 feet in the air and directed towards a street. Sample data is provided for each of the cameras and includes one full day of sub-sampled data. The sampled data consists of approximately 1,000 to 1,500 images for each camera. As of now, there are 3 to 6 months of data available for each camera. Larger datasets can be requested via the form below. The numbers in the image correspond to hte camera # links below. Please specify "Suburb Dataset" with the camera number when submitting a request.

Funded in part by DOT grant #69A3551747111 entitled "Better Boulevard Analytics".

Download Links: [camera 1] (2.58 GB) - [camera 2] (2.4 GB) - [camera 3] (1.51 GB) - [camera 4] (1.33 GB) - [camera 5] (1.37 GB) - [camera 6] (1.17 GB)

Larger Data Access Request

We have released a sample data from mutiple cameras in the links above. For larger dataset please fill the form to request for additional data. Please note that the dataset is humongous and releasing the complete dataset is not feasable. So be reasonable in your request and we will try to get the data as soon as possible.

Copyright