This is the food52 fork of thumbor.
We've made the following changes:
-
For performance, don't check the results storage for each request. Our results storage is S3, so these checks are very expensive, and since our infrastructure tries to route resizing requests to thumbor only when they don't exist in S3, there's little upside.
-
Make a signal handler reentrant.
-
Add mechanism for graceful shutdown.
-
Add option to specify logfile path.
Our production system is currently branched of off thumbor version 4.11.1;
the ref is 4.11.1-f52v1. Our development is on food52-mainline, which is
the default branch in GitHub.
Version 5 appears to have involved a major refactoring, so there will probably be some pain in merging our changes with upstream
thumbor is a smart imaging service. It enables on-demand crop, resizing and flipping of images.
It also features a VERY smart detection of important points in the image for better cropping and resizing, using state-of-the-art face and feature detection algorithms (more on that in Detection Algorithms).
Using thumbor is very easy (after it is running). All you have to do is access it using an URL for an image, like this:
http://<thumbor-server>/300x200/smart/s.glbimg.com/et/bb/f/original/2011/03/24/VN0JiwzmOw0b0lg.jpg
That URL would show an image of the big brother brasil participants in 300x200 using smart crop.
There are several other options to the image URL configuration. You can check them in the Usage page.
For more information check thumbor's documentation.
You can see thumbor in action at http://thumborize.me/