The Boston Timescape Dataset

contact: Adrian Dalca

mouse over the banner!

1M Boston Dataset

This page documents ~0.9 million pictures of the Boston skyline (and ~0.1 million of Cambridge) taken over ~5 years from a height of about 55 meters up.

The database currently has on the order of 0.91 million images (JPG, sometimes bracketed for HDR) with 11,000 also having RAW formats. They are taken with various cameras: a Canon 5D Mark II (21 MP), a GoPro Hero 2, a GoPro Hero 3, and several point and shoots/phone cameras. Some videos are also included. A very rough estimate of the distribution of the images through time (the colors indicate seasons) shows many more images recently compared to 2010. Some pictures are singletons (just one picture taken for in a few days), but many are part of timelapses (pictures taken ~every second or minute).

The database is currently not curated (there are some inconsistencies) and, most importantly, the images are not precisely aligned.

Data Sample

Below, I've sampled a few untouched images, randomly (roughly uniformly over time periods). This sample will be roughly indicative of the quality of the data in terms of camera angle, lighting, etc - including pictures that don't belong but were included by accident. The images have been rescaled to 540 x 960px. On the right, we curated and aligned a small sample.


Preview 264 images: move mouse left-right over images.

grid view (dropbox)
tipiX View

Aligned and Selected

Here's a tipiX View of 21 aligned images.

The images are more heterogeneous earlier in the dataset. Often I would attempt panorama shots, so some of the randomly sampled images are not at the approximate same angle as the rest of the images. Some images might not be landscape orientation, might have reflections, or different zoom levels. All of this can be cleaned up with some semi-manual work :).


I've formed several panoramas of Boston + Cambridge (I think this covers about 240 degrees), here are a couple of examples:


I tended to take more pictures at interesting times - interesting weather, clouds, moon, etc. You can also spot some Boston events - such as the transistor fire in 2012 (news story), the Dedham brush fire of 2012 (news story) or July 4th firworks.

Timelapse in a photo

The dataset includes several blocks of images taken a few seconds apart for creating timelapse videos. Combining these frames into a single image, where each vertical line is taken from a different consecutive frame, shows interesting daily motions.


Here are some season averages of a small set of non-aligned, non-curated images


If this dataset sounds interesting in any way, please contact me and I'm more than happy to discuss and share. I'm currently trying to clean up and organize the nearly 3TB of data.