elyha7 / yoloface Goto Github PK

Hello, mate @elyha7 .
I'm trying to convert face detector to onnx via your "export.py" file and have some issue:
after convertion output sizes don't fit with output sizes from loaded model ("model = YoloDetector( ... )")

Is it ok? And I'm confused, which postprocessing I need to do with these outputs? Any ideas?
P.S. thx a lot for your repo))

Hi,
I can not use your Repo for inference.

I have 'cloned' your repo, 'cd' to it, then installed dependencies which has been mentioned in 'requirements.txt'
and then used

from yoloface import YoloDetector
import numpy as np
from PIL import Image

model = YoloFace(target_size=720,gpu=0,min_face=90)
orgimg = np.array(Image.open('test_image.jpg'))
bboxes,points = model.predict(orgimg)

But it returns that yoloface is not installed!

---

ImportError Traceback (most recent call last)
in ()
----> 1 from yoloface import YoloDetector
2 import numpy as np
3 from PIL import Image
4
5 model = YoloFace(target_size=720,gpu=0,min_face=90)

/content/yoloface/yoloface.py in ()
11 from math import sqrt
12
---> 13 from .models.common import Conv
14 from .models.yolo import Model
15 from .utils.datasets import letterbox

ImportError: attempted relative import with no known parent package

---

NOTE: If your import is failing due to a missing package, you can
manually install dependencies using either !pip or !apt.

---

How should i fix this?

Thanks
Best regards

Where can I find weight for different models? yolov5n-face, yolov5s-face, yolov5m-face and so on?

How do I choose a model?

I am trying to use your face detector script to get the bboxes on several images but it does not work how it is expected. First of all, your README describes that you can pass several images packed in a list to get multi-image predictions with the following example:
bboxes,points = model.predict([image1,image2])
but does not mention if the image1 and image2 are the paths to that images or the object you get after you open them with PIL (for example: Image.open('image1'), assuming they have the same root directory). I get the following errors in this cases:

Traceback (most recent call last):
File "C:\YOLOv5\yoloface\detect.py", line 65, in
_main(arguments)
File "C:\YOLOv5\yoloface\detect.py", line 58, in _main
bboxes, points = model.predict(orgimg)
File "C:\YOLOv5\yoloface\face_detector.py", line 159, in predict
images = self._preprocess(images)
File "C:\YOLOv5\yoloface\face_detector.py", line 73, in _preprocess
h0, w0 = img.shape[:2] # orig hw
ValueError: not enough values to unpack (expected 2, got 1)

from this call: orgimg = np.array([Image.open('00000002.png'), Image.open('00000001.png')]) bboxes, points = model.predict(orgimg) (when I pass the images as PIL objects)

Traceback (most recent call last):
File "C:\YOLOv5\yoloface\detect.py", line 65, in
_main(arguments)
File "C:\YOLOv5\yoloface\detect.py", line 58, in _main
bboxes, points = model.predict(orgimg)
File "C:\YOLOv5\yoloface\face_detector.py", line 159, in predict
images = self._preprocess(images)
File "C:\YOLOv5\yoloface\face_detector.py", line 73, in _preprocess
h0, w0 = img.shape[:2] # orig hw
ValueError: not enough values to unpack (expected 2, got 1)

from this call: orgimg = np.array(['00000002.png', '00000001.png']) bboxes, points = model.predict(orgimg) (where I pass the images as paths to them)

Traceback (most recent call last):
File "C:\YOLOv5\yoloface\detect.py", line 65, in
_main(arguments)
File "C:\YOLOv5\yoloface\detect.py", line 58, in _main
bboxes, points = model.predict(['00000002.png', '00000001.png'])
File "C:\YOLOv5\yoloface\face_detector.py", line 159, in predict
images = self._preprocess(images)
File "C:\YOLOv5\yoloface\face_detector.py", line 73, in _preprocess
h0, w0 = img.shape[:2] # orig hw
AttributeError: 'str' object has no attribute 'shape'

from this call: bboxes, points = model.predict(['00000002.png', '00000001.png']) (where I pass the images as paths to them)

and

Traceback (most recent call last):
File "C:\YOLOv5\yoloface\detect.py", line 65, in
_main(arguments)
File "C:\YOLOv5\yoloface\detect.py", line 58, in _main
bboxes, points = model.predict([Image.open('00000002.png'), Image.open('00000001.png')])
File "C:\YOLOv5\yoloface\face_detector.py", line 159, in predict
images = self._preprocess(images)
File "C:\YOLOv5\yoloface\face_detector.py", line 73, in _preprocess
h0, w0 = img.shape[:2] # orig hw
File "C:\Users\rtanase\Anaconda3\envs\detect\lib\site-packages\PIL\Image.py", line 519, in getattr
raise AttributeError(name)
AttributeError: shape

from this call: bboxes, points = model.predict([Image.open('00000002.png'), Image.open('00000001.png')]) (when I pass the images as PIL objects)

Hi!
Firstly, thank you so much for such a convenient wrapper, you saved a lot of time, thank you again!
When trying to "drive away" a vertical photo, unfortunately I encounter an error

RuntimeError: Given groups=1, weight of size [32, 3, 3, 3], expected input[1, 4, 608, 608] to have 3 channels, but got 4 channels instead

I will be glad to help, thanks a lot in advance!
Have a nice day :D

I think this needs to be
sys.path.append(os.path.abspath(os.path.dirname(__file__)))
instead of
sys.path.append(os.path.abspath(os.path.join(os.path.dirname("__file__"), '..')))

Hello!
In source code there is mentioned anti profile model to determine, whether a face is frontal or not

But there is no such file in models folder, nor the mentioning about that in readme

Can you please tell me, what is it and where could it be found?

Where did you find these coordinates?
thank you

In README you use as an example YoloDetector(target_size=720,gpu=0,min_face=90). But YoloDetector has no parameter gpu, i guess it should be device

Can't get attribute 'SPPF' on <module 'models.common' from
when model = torch.load('yolov5m.pt', map_location='cpu')['model']

Has anyone compared SCRFD to YOLO for face detection? Which one is better or more performant?

Hello,

Is there any support for training / fine-tune weights on personal datasets? Or I should use the train script from the deepcam repository? Thank you in advance.