top | item 42816472

(no title)

DarkSucker | 1 year ago

Can anyone speak to the image processing used? In particular, how is depth inferred from a single 2D image? It seems to me one would need both depth and angle over the field of view to back out the lens focal length. The EXIF format doesn't seem to contain meta data helpful to the focal length calculation.

discuss

order

kan02134|1 year ago

As @nanoanderson mentioned, we use the "FocalLengthIn35mmFormat" tag from EXIF. Let me explain why: since cameras have different sensor sizes, comparing actual focal lengths directly can be misleading (https://en.wikipedia.org/wiki/Crop_factor#Introduction). By using the 35mm film equivalent, we can compare field of views across all cameras on the same scale. For instance, a 50mm lens on a crop sensor camera shows a similar field of view to a 75mm lens on a full-frame camera.

We're using the 'exifr' library for EXIF data extraction: https://www.npmjs.com/package/exifr