Simswap not works

[nascacorp]

Getting an error with two Simswap collab (pictures/videos)

[woctezuma]

What is the error message?

Edit: Copied below: ValueError: This ORT build has...

Traceback (most recent call last):
  File "/content/SimSwap/test_wholeimage_swapsingle.py", line 55, in <module>
    app = Face_detect_crop(name='antelope', root='./insightface_func/models')
  File "/content/SimSwap/insightface_func/face_detect_crop_single.py", line 40, in __init__
    model = model_zoo.get_model(onnx_file)
  File "/usr/local/lib/python3.10/dist-packages/insightface/model_zoo/model_zoo.py", line 56, in get_model
    model = router.get_model()
  File "/usr/local/lib/python3.10/dist-packages/insightface/model_zoo/model_zoo.py", line 23, in get_model
    session = onnxruntime.InferenceSession(self.onnx_file, None)
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 432, in __init__
    raise e
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 419, in __init__
    self._create_inference_session(providers, provider_options, disabled_optimizers)
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 451, in _create_inference_session
    raise ValueError(
ValueError: This ORT build has ['AzureExecutionProvider', 'CPUExecutionProvider'] enabled. Since ORT 1.9, you are required to explicitly set the providers parameter when instantiating InferenceSession. For example, onnxruntime.InferenceSession(..., providers=['AzureExecutionProvider', 'CPUExecutionProvider'], ...)

Traceback (most recent call last):
  File "/content/SimSwap/test_wholeimage_swapmulti.py", line 56, in <module>
    app = Face_detect_crop(name='antelope', root='./insightface_func/models')
  File "/content/SimSwap/insightface_func/face_detect_crop_multi.py", line 40, in __init__
    model = model_zoo.get_model(onnx_file)
  File "/usr/local/lib/python3.10/dist-packages/insightface/model_zoo/model_zoo.py", line 56, in get_model
    model = router.get_model()
  File "/usr/local/lib/python3.10/dist-packages/insightface/model_zoo/model_zoo.py", line 23, in get_model
    session = onnxruntime.InferenceSession(self.onnx_file, None)
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 432, in __init__
    raise e
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 419, in __init__
    self._create_inference_session(providers, provider_options, disabled_optimizers)
  File "/usr/local/lib/python3.10/dist-packages/onnxruntime/capi/onnxruntime_inference_collection.py", line 451, in _create_inference_session
    raise ValueError(
ValueError: This ORT build has ['AzureExecutionProvider', 'CPUExecutionProvider'] enabled. Since ORT 1.9, you are required to explicitly set the providers parameter when instantiating InferenceSession. For example, onnxruntime.InferenceSession(..., providers=['AzureExecutionProvider', 'CPUExecutionProvider'], ...)

[woctezuma]

A priori, it is easy to fix, but it should be done in the official repositories of SimSwap or insightface.

Related:

• Problem ValueError with SimSwap neuralchen/SimSwap#445
• ValueError: This ORT build has ['TensorrtExecutionProvider', 'CUDAExecutionProvider', 'CPUExecutionProvider'] enabled. neuralchen/SimSwap#316 (comment)

[woctezuma]

I can confirm that editing the following file inside insightface works.

File "/usr/local/lib/python3.10/dist-packages/insightface/model_zoo/model_zoo.py", line 23

    def get_model(self):
-        session = onnxruntime.InferenceSession(self.onnx_file, None)
+        session = onnxruntime.InferenceSession(self.onnx_file, None, providers=['AzureExecutionProvider', 'CPUExecutionProvider'])

[woctezuma]

!pip install -U insightface

Upgrading insightface works as well, but leads to another issue with SimSwap.

Traceback (most recent call last):
  File "/content/SimSwap/test_wholeimage_swapsingle.py", line 62, in <module>
    img_a_align_crop, _ = app.get(img_a_whole,crop_size)
  File "/content/SimSwap/insightface_func/face_detect_crop_single.py", line 64, in get
    bboxes, kpss = self.det_model.detect(img,
TypeError: RetinaFace.detect() got an unexpected keyword argument 'threshold'

• THRESHOLD ERROR neuralchen/SimSwap#139

[woctezuma]

To summarize, the following happens with insightface==0.6.2 or below.

ValueError: This ORT build has ['AzureExecutionProvider', 'CPUExecutionProvider'] enabled. Since ORT 1.9, you are required to explicitly set the providers parameter when instantiating InferenceSession. For example, onnxruntime.InferenceSession(..., providers=['AzureExecutionProvider', 'CPUExecutionProvider'], ...)

---

The following happens with insightface==0.7 or above.

TypeError: RetinaFace.detect() got an unexpected keyword argument 'threshold'

---

The version history is at https://pypi.org/project/insightface/#history

[woctezuma]

Fixed for now with a fork of SimSwap which removes the threshold for Retinaface.

• 074a71b Upgrade insightface to version 0.7.3
• 21f45b2 Rely on my fork of SimSwap

See that there is indeed no 'threshold' argument:

class RetinaFace:
    def detect(self, img, input_size = None, max_num=0, metric='default'):

[nascacorp]

Working now! thanks a lot pal

[woctezuma]

You are welcome! Thank you for reporting this issue!

---

Regarding my solution, I had the choice between fixing:

• ValueError: This ORT build has...,
• TypeError: RetinaFace.detect() got ....

I went with the latter and I believe there should be no issue with this solution. Indeed:

1. the default value of self.det_thresh is the same (0.5):

class RetinaFace:
    def __init__(self, model_file=None, session=None):
        self.det_thresh = 0.5

vs. inside SimSwap:

class Face_detect_crop:
    def prepare(self, ctx_id, det_thresh=0.5, det_size=(640, 640), mode ='None'):
        self.det_thresh = det_thresh

2. self.det_thresh is directly used inside detect():

    def detect(self, img, input_size = None, max_num=0, metric='default'):
        scores_list, bboxes_list, kpss_list = self.forward(det_img, self.det_thresh)

without the need to specify the variable in the function arguments:

    def get(self, img, crop_size, max_num=0):
        bboxes, kpss = self.det_model.detect(img,
-                                            threshold=self.det_thresh,
                                             max_num=max_num,
                                             metric='default')

However:

1. In practice, SimSwap uses the value of 0.6 rather than the default value of 0.5.

    app = Face_detect_crop(name='antelope', root='./insightface_func/models')
    app.prepare(ctx_id= 0, det_thresh=0.6, det_size=(640,640),mode=mode)

2. I also have to set self.det_model.det_thresh instead of self.det_thresh.

    def prepare(self, ctx_id, det_thresh=0.5, det_size=(640, 640), mode ='None'):
-        self.det_thresh = det_thresh
+        self.det_model.det_thresh = det_thresh

Edit: Done! And I have sent a pull request to the official repository:

• Upgrade insightface neuralchen/SimSwap#447