Closed bsormagec closed 7 months ago
Is ok for me, what is you version and requests?
Latest version and use different size mask and image.
Confirmed. fix it later.
only one line changed, you can have a try.
work.py, line 412
inpaint_mask = np.maximum(inpaint_mask, inpaint_mask_image_upload)
to
inpaint_mask = np.maximum(np.zeros(shape=(W, H), dtype=np.uint8), inpaint_mask_image_upload)
This is because in Fooocus webui , when you choice upload mask, app will generate a empty mask and merge with resized mask you upload. It uses two variables named inpaint_mask
and inpaint_mask_image_upload
to represent the two mask
Since version 0.3.29 in Fooocus-API, The logic of worker has changed, and we have no inpaint_mask_image_upload
, It is generated by the program automatically resize by inpaint_mask
in worker.py
, As a result, there will be size inconsistencies in the final merger.
Thank you my friend.
now have same issue on input_image. :D trying to find which line is this.
Worker error: operands could not be broadcast together with shapes (1440,1800) (1800,1440)
I fixed it on https://github.com/mrhan1993/Fooocus-API/pull/300 @mrhan1993 merge it then it works.
now have same issue on input_image. :D trying to find which line is this.
Worker error: operands could not be broadcast together with shapes (1440,1800) (1800,1440)
I fixed it on #300 @mrhan1993 merge it then it works.
My mistake :joy:
When I try to use /image-inpaint-outpaint endpoint with mask, it returns that error. but it works on fooocus, broken on API only did you skip something ? @mrhan1993