3 ms·
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 o
by DarkSucker 2y 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.
- nanoanderson 2y agoThere is a FocalLength tag in the EXIF spec. https://exiftool.org/TagNames/EXIF.html https://exiftool.org/TagNames/EXIF.html
- kan02134 2y agoAs @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 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 https://www.npmjs.com/package/exifr