One of our assemblies carries out several resize steps to create various sized images from the original. One of the sizes we have the most problems with is a 900x400 header style image. Quite often the crop (as part of the resize robot) takes the heads off people in the photo. This is more pronounced when a portrait photo is uploaded.
Is there a way to use the facedetect robot to provide an optimal crop for the resize robot. Perhaps setting the ‘gravity’ attribute to whichever would provide the optimal face inclusion? I think that cloudinary offers something similar: https://cloudinary.com/blog/face_detection_based_cropping
Any suggestions gratefully received.
P.S We are implementing a manual cropping tool for manually adjusting individual images but would like the automated process to be a little more effective.
We don’t support face based cropping on our /image/resize robot for the moment. But one you can go about it, is to implement it using the facedetect robot. Checking robot params https://transloadit.com/docs/transcoding/#image-facedetect . You can see that you have crop_padding that can center on the face and extend to a specific height/width from there. It can even use a percentage there.
How does that sound to you?
Thanks, I’ll give that a go as an extra step.
I have had a quick go using the facedetect robot and it seems rather limited. I am testing using a portrait photo of 2 people. The robot only seems to pick up 1 of the faces and therefore when I alter the crop padding setting the photo is unbalanced as it has focussed on only one person. Happy to share photo and template settings via the main account support channel.
Yes, please. Can you please send me your setup on firstname.lastname@example.org or on our support channel. I will be more than happy to take a look at it.
Have sent info through via intercom
Hi all, just for googlers stumbling upon this thread, did we solve this, and if so, how?
No, unfortunately it wasn’t solved. As it stands, the image resize robot and the facedetect robot cannot be used together to create a cropped image that focuses on the people in the photo.
Something for the future I hope.
Ah okay, I’ll make a note of it and see if we can deliver this soon!