I was wondering weather is it possible to use lighter models such as mobilenet as a replacement for convolutional stack in this example? I am confused in the step where the convolutional and GRU layers combined.
conv_to_rnn_dims = (img_w // (pool_size ** 2), (img_h // (pool_size ** 2)) * conv_filters) inner = Reshape(target_shape=conv_to_rnn_dims, name='reshape')(inner) inner = Dense(time_dense_size, activation=act, name='dense1')(inner) gru_1 = GRU(rnn_size, return_sequences=True, kernel_initializer='he_normal', name='gru1')(inner) gru_1b = GRU(rnn_size, return_sequences=True, go_backwards=True, kernel_initializer='he_normal', name='gru1_b')(inner) ... ...
Any idea about this?
TIA
I was wondering weather is it possible to use lighter models such as mobilenet as a replacement for convolutional stack in this example? I am confused in the step where the convolutional and GRU layers combined.
conv_to_rnn_dims = (img_w // (pool_size ** 2), (img_h // (pool_size ** 2)) * conv_filters) inner = Reshape(target_shape=conv_to_rnn_dims, name='reshape')(inner) inner = Dense(time_dense_size, activation=act, name='dense1')(inner) gru_1 = GRU(rnn_size, return_sequences=True, kernel_initializer='he_normal', name='gru1')(inner) gru_1b = GRU(rnn_size, return_sequences=True, go_backwards=True, kernel_initializer='he_normal', name='gru1_b')(inner) ... ...Any idea about this?
TIA