mjdelta / face-recognition-keras Goto Github PK

In your "test.py" file,
12 model_path=BASE_MODEL+".h5"

However, I cannot find the vgg16.h5 in your repository.

Plz Upload them.

Hello,
I downloaded the LFW dataset, and placed it in the directory, and when I run train.py I get this error:

File "/scratch/gits/face-recognition-keras/data_utils.py", line 100, in flow
ans,pos,nes,anchors_names,positives_names,negatives_names=self.get_triplet(values,names)
File "/scratch/gits/face-recognition-keras/data_utils.py", line 74, in get_triplet
choosed_faces_values=choosed_faces_values.reshape((len(choosed_faces_values),96,96,3))
ValueError: cannot reshape array of size 431812500 into shape (2303,96,96,3)

Is there a gudie what tha training data looks like in the directory?
And I want to train this on other topic.
But it's the same as the triplet method.
But I got the memory error in the training.

RuntimeError Traceback (most recent call last)
in ()
10 img_path=os.path.join(dir_,d,f)
11 size=96
---> 12 img=align_face(img_path,size)
13 if np.sum(img)==-1:
14 continue

in align_face(img_path, size)
4 if np.sum(bb)==-1:
5 return -1
----> 6 landmarks=get_face_shape(img,bb)
7
8 landmarksindices=np.array(OUTER_EYES_AND_NOSE)

in get_face_shape(img, bb)
11 def get_face_shape(img,bb):##return 68 keypoints of face
12 model_path="models/landmarks.dat"
---> 13 predictor=dlib.shape_predictor(model_path)
14 pnts=predictor(img,bb)
15 return np.float32(list(map(lambda p:(p.x,p.y),pnts.parts())))

RuntimeError: Unable to open models/landmarks.dat