I am using google colab to train a model for sign language recognition using this dataset: https://www.kaggle.com/feronial/turkish-sign-languagefinger-spelling
I tried different architecture for my CNN model but this one gives me the best results so far:
l2 = regularizers.l2(0.001) model = Sequential([ Conv2D(16, 3, padding='same', activation='relu',kernel_regularizer=l2, input_shape=(224, 224 ,3)), MaxPooling2D(), Dropout(0.35), Conv2D(32, 3, padding='same', activation='relu',kernel_regularizer=l2), MaxPooling2D(), Dropout(0.35), Conv2D(64, 3, padding='same', activation='relu',kernel_regularizer=l2), MaxPooling2D(), Dropout(0.35), Conv2D(128, 3, padding='same', activation='relu',kernel_regularizer=l2), MaxPooling2D(), Dropout(0.35), Flatten(), Dense(1024, activation='relu',kernel_regularizer=l2), Dropout(0.35), Dense(512, activation='relu',kernel_regularizer=l2), Dropout(0.35), Dense(23,activation="softmax") ])
i am using an ADAM optimizer with lr=0.001 and batch size of 32
i tried training for 50,100,200 epochs but the results weren't so much different.
I am getting 99-100% accuracy on training,
but only 70-75% accuracy on validation,
and a very low 30-40% accuracy on test.
can anyone point in a direction to improve these numbers?
I would recommend you to use https://www.comet.ml/site/ website.
You can easily do hyperparameter tuning by it.
Hi
Please check the following link. It describes the process in details. I believe it will help you.
https://ai.stackexchange.com/questions/8414/how-to-improve-testing-accuracy-when-training-accuracy-is-high
Regards
Narges Takhtkeshha I have integrated this site with my code but I wasn't able to do hyperparameter tuning with it, can you provide any help for using this site?
Ammar Albakri You should do these steps:
1- Build a new project in comet.ml site
2-copy your given experiment statement (at the bottom of the page)
for example
experiment = Experiment(api_key="MrwDCgaK0BqzgRHv04J268iS1", project_name="mnist", workspace="Narges Takhtkeshha")
3-Afterwards in google colab or any programming environment do such a thing:
experiment = Experiment(api_key="MrwDCgaK0BqzgRHv04J268iS1", project_name="mnist", workspace="Narges Takhtkeshha")
#Set your hyperparameter
learning_rate = 0.1
epochs = 20 batch_size = 32
act_fn = 'relu'
h_layer_1 = 20
h_layer_2 = 20
params={'batch_size': batch_size, epochs':epochs, 'layer_activation':act_fn, 'optimizer':'adam', 'learning_rate': learning_rate, "h_layer_1" :h_layer_1, "h_layer_2" :h_layer_2, }
# log parameters in Comet.ml
experiment.log_parameters(params)
- Badges
- Science topic
- Similar topics
- Organizational Psychology
- Training