Load facenet model

Question

I have tried almost all the answers on stackoverflow but nothing worked. Here is my code.

from keras.models import load_model
load_model('facenet_keras.h5')

It is giving me this error

ValueError Traceback (most recent call last) ~\AppData\Local\Temp\ipykernel_5776\2622147163.py in ----> 1 load_model('facenet_keras.h5')

~\AppData\Roaming\Python\Python39\site-packages\keras\utils\traceback_utils.py in error_handler(*args, **kwargs) 68 # To get the full stack trace, call: 69 # tf.debugging.disable_traceback_filtering() ---> 70 raise e.with_traceback(filtered_tb) from None 71 finally: 72 del filtered_tb

~\AppData\Roaming\Python\Python39\site-packages\keras\utils\generic_utils.py in func_load(code, defaults, closure, globs) 101 except (UnicodeEncodeError, binascii.Error): 102 raw_code = code.encode("raw_unicode_escape") --> 103 code = marshal.loads(raw_code) 104 if globs is None: 105 globs = globals()

ValueError: bad marshal data (unknown type code)

To solve the above error I did this

from keras_facenet import FaceNet
embedder = FaceNet()

But I don't want to use the above method.I want to load the facenet model only.How to solve this error if anyone can help.

1. Python verison : 3.9.3
2. tensorflow : 2.11.0
3. keras : 2.11.0

EDIT

According to V.M's answer, this worked.

model = InceptionResNetV1(
        input_shape=(None, None, 3),
        classes=512,
    )
model.load_weights('20180402-114759.h5')

Answer 1

If you can recreate the architecture, in this case from [keras_facenet/inception_resnet_v1][1], then you can do:

model = InceptionResNetV1(
        input_shape=(None, None, 3),
        classes=512,
    )
model.load_weights('model.h5')

Answer 2

Use this insted:

model = tf.keras.applications.ResNet50(weights='imagenet')

This works for me. In our code, you will not be able to use the 'predict' attribute for embeddings, but using this you can.

before implementing you just need to install one dependency using the bellow code in your terminal:

#pip install keras_facenet