Dataset with labels
Image filename column
Label column
Train ratio
Proportion of the sample that goes to the train set. The rest goes to the test set
Random Seed
Using a fixed random seed allows for reproducible result
Model
Pooling
Image width/height
Input model has already be retrained. Pooling and size are already set.
Layer(s) to retrain
Dropout
L1 regularization
L2 regularization
{{getShowHideAdvancedParamsMessage()}}
{{modelSummary}}
Optimization
Optimizer
Learning rate
Custom Parameters
Add
Training
Batch size
Number of epochs
Steps per epoch
Number of validation steps
Data Augmentation
# augmentation per image
Data augment. custom params
Add
Tensorboard
You can access tensorboard via a DSS webapp
GPU
Use GPU
List of GPUs to use
Comma separated list of GPU indexes
Memory allocation rate per GPU