FaceRestoreHelper seems not thread-safe

[henryruhs]

Using FaceRestoreHelper needs with threading.Semaphore() to work within threads. I experienced this first in GFPGAN and later in our custom implementations.

This limits performance of GFPGAN a lot!

[I8Robot]

this error:

"facexlib\utils\face_restoration_helper.py", line 151, in get_face_landmarks_5
    self.det_faces.append(bbox[0:5])
AttributeError: 'numpy.ndarray' object has no attribute 'append'"?

May I ask where to add with threading.Semaphore()? I also want to work with multithreads.

[henryruhs]

Just before using it, you can create the instance via threading.Lock() before.
You can find our code (using gfpgan) here: https://github.com/s0md3v/roop/blob/main/roop/processors/frame/face_enhancer.py

[I8Robot]

Just before using it, you can create the instance via threading.Lock() before. You can find our code (using gfpgan) here: https://github.com/s0md3v/roop/blob/main/roop/processors/frame/face_enhancer.py

OK, roop 😃