keras-arcface
keras-arcface copied to clipboard
Keras implementation of ArcFace, CosFace, and SphereFace
can any one show me how to input image dataset to that model and how to get results please
Could you please look at my code, if I am using the model correctly? y = Input(shape=(10,)) mobile_V2 = MobileNet(input_shape =(32,32,3), include_top = False, alpha = 0.5) input = mobile_V2.input...
And if the parameter m is set like 4, the label must be leaked, because the training accuracy, validation accracy and test accuracy are all 1.0 even in the 1st...
`num_classes` to `n_classes`
Hello! I am trying to use arcface loss with mobilenet_v3,but I got above question,can u help me to fix this out?Thank u! ``` def build(self, input_shape): super(ArcFace, self).build(input_shape[0]) self.W =...
https://github.com/4uiiurz1/keras-arcface/blob/877ea00531d74c735354a6cee92788110542196e/metrics.py#L36 at this point the range of theta is `0
Is it possible to change margin while training? For example increment it for 0.01 every epoch?
I have tried this code, but it's doesn't training at all.
NotImplementedError: Layer ArcFace was created by passing non-serializable argument values in `__init__()`, and therefore the layer must override `get_config()` in order to be serializable. Please implement `get_config()`. Example: class CustomLayer(keras.layers.Layer):...
How can we predict the probability of test samples without passing the ground truth labels during model.predict(). I think passing the labels will defeat the purpose of blind testing. Please...