webp image format support

Feature Requests
1

Platform(s), such as mobile, website, API, other: All

URLs (aka web addresses) of any pages, if relevant:

Description of need:
Support the .webp image format for image upload.

Feature request details:
The webp image format is becoming more and more widely used due to its characteristics superior to other (older) image formats, like the image quality and smaller size. As a container, besides lossless compression for static images, it also supports animated images and videos.

Although there is a topic https://forum.inaturalist.org/t/webp-image-support/18833 named “webp image format”, it strangely shifts to a request for video support and has been denied, due to that fact, apparently.

The newer camera apps are generating webp images, which are significantly smaller than JPGs and PNG, while retaining the full image quality, but I am prevented from uploading these to iNaturalist either on the mobile (Android) or the web.

2

is this actually true for most photos uploaded to iNat? see: https://siipo.la/blog/is-webp-really-better-than-jpeg.

5 Likes
3

From my own testing I would say generally no. It is highly dependent on the nature of the image in question; occasionally it can be noticeably better than JPG, and sometimes even a little worse. But for most wildlife style images rarely worthwhile I found. iNaturalist have done some of their own testing on this – see end of tiwane’s post here:
https://forum.inaturalist.org/t/increase-max-image-size-from-2048-x-2048-pixels/13627/26

Though at the same time, I don’t know if there’s any good reason not to support it?

4

That article mentions an aspect of WebP that could be problematic for iNat images:

JPEG compression uses the same quantization factor for each 16×16 “macroblock” so the compression is consistent throughout the image.

WebP and AVIF on the other hand use different compression factors for different parts of the image so while the detailed parts of the image retain their quality, surfaces like skin or the sky which have low detail are “smoothed out”.

I’m concerned that an algorithm that handles compression differently in different parts of the image might obscure details that could be important for a human identifier or the computer vision algorithm.

There’s a broader issue here for CV, too: Would a CV model trained on JPEG images perform equally well at identifying images encoded with the WebP algorithm?

1 Like