AMOS: Dataset Details

return to AMOS overview.

This page describes the organization of the AMOS dataset. If this page does not contain the information you need let us know.

Obtaining the Dataset

We are just starting to share the dataset and we are still working on the protocol for obtaining the dataset. Two options that have worked well are:
  1. If you want images from a few cameras: We can send you a DVD-R. You just need to select the cameras of interest (possibly by browsing the scene archives).
  2. If you want all of the data (around 400GB on 7/19/2007), or a large subset, you should send us a FAT32-formatted USB drive and we will send it back loaded with data.
Please contact Nathan Jacobs if you would like access to the AMOS dataset.

Dataset Organization

All images from a given camera are grouped in a single directory with file names (described below) that correspond to when the image was captured.

/{Camera ID}/image files

Camera Metadata

If there is interest we can make available the geolocation and a textual description of the scene of many of the cameras in the dataset.

Image File Names

Filenames are determined by the download time of the file. The download time is not always the image capture time (some images are delayed because they are uploaded to a webserver).

The image filename format is : yyymmdd_HHMMSS.jpg (using Matlab date syntax).

We use the following Matlab code to load the files from a given camera (in the cBase directory) and to extract the date from the filename:

D = dir([cBase '20*.jpg']); 
D = D(0 < [D.bytes]); % remove empty files
dateNums = datenum({D.name}, 'yyyymmdd_HHMMSS');

All times are recorded in Central Time (UTC-5 or UTC-6 depending on the time of year). We use the following Matlab script to remove daylight savings time.

% convert times with daylight savings shift to standard time
function d = cdt2cst(d)

  ds = datenum({'April 6, 2003', 'April 4, 2004','April 3, 2005', ...
    'April 2,2006', 'March 11, 2007', 'March 9, 2008', ...
    'March 8, 2009', 'March 14, 2010'});

  de = datenum({'Oct 26, 2003', 'Oct 31, 2004','Oct 30, 2005', ...
    'Oct 29,2006', 'November 4, 2007', 'Nov 2, 2008', ...
    'Nov 1, 2009', 'Nov 7, 2010'});

  for ix = 1:numel(de)
    ind = (ds(ix)<=d) & (d < de(ix));
    d(ind) = d(ind) - 1/24;
  end

end