Save models

[InstantWindy]

I have a question, after training the model ,how to save the model's weight? For every scene,I need to save corresponding weight,?if so ,it needs to save 52 weight files, could you help me?

[lim-anggun]

Hi @InstantWindy, it a scene-specific method as described in the paper. To save only weights, set save_weights_only=True To save only models, set save_weights_only=False

chk = keras.callbacks.ModelCheckpoint(mdl_path, monitor='val_loss', verbose=0, save_best_only=True, save_weights_only=True, mode='auto', period=1)

[InstantWindy]

but if set save_weights_only=True,where the weight file is? For every scene ,after training the model is deleted, when testing ,how to use the trained model? I try to use the code "model.save_weights('weights.h5') " for every scene saving the weights. I don't know, could you explain in detail?Thank you

[lim-anggun]

@InstantWindy, the following Keras function saves the model for you automatically. Each model is saved at mdl_path, where mdl_path is the path of your model: eg:mdl_path = '/models/my_highway_model.h5'

chk = keras.callbacks.ModelCheckpoint(mdl_path, monitor='val_loss', verbose=0, save_best_only=True, save_weights_only=False, mode='auto', period=1)

after training, load the model from mdl_path:

from keras.models import load_model
model = load_model(mdl_path)
model.predict(...)

There are some differences between keras model and weight, you can read Keras docs for more details. Here

[LavenderMP]

After i trained the whole CDnet2014 i got the .h5 model but i can't load it it gives erros: Traceback (most recent call last): File "test.py", line 6, in <module> model = load_model(PATH) File "C:\Users\hvhnk\Anaconda3\envs\tf-gpu\lib\site-packages\keras\models.py", line 264, in load_model sample_weight_mode=sample_weight_mode) File "C:\Users\hvhnk\Anaconda3\envs\tf-gpu\lib\site-packages\keras\engine\training.py", line 681, in compile loss_function = losses.get(loss) File "C:\Users\hvhnk\Anaconda3\envs\tf-gpu\lib\site-packages\keras\losses.py", line 102, in get return deserialize(identifier) File "C:\Users\hvhnk\Anaconda3\envs\tf-gpu\lib\site-packages\keras\losses.py", line 94, in deserialize printable_module_name='loss function') File "C:\Users\hvhnk\Anaconda3\envs\tf-gpu\lib\site-packages\keras\utils\generic_utils.py", line 159, in deserialize_keras_object ':' + function_name) ValueError: Unknown loss function:loss

[lim-anggun]

Hi @LavenderMP, You need to add custom objects when loading the model:

from my_upsampling_2d import MyUpSampling2D
from FgSegNet_M_S_module import loss, acc, loss2, acc2

model = load_model(model_path, custom_objects={'MyUpSampling2D': MyUpSampling2D,  'loss':loss, 'acc':acc, 'loss2':loss2, 'acc2':acc2})