Train Binary CNN Models

This function trains Convolutional Neural Network (CNN) models, such as AlexNet, VGG16, VGG19, ResNet18, ResNet50, or ResNet152, on a given dataset. The trained model is saved along with metadata for further usage.

train_CNN_binary(
  input.data.path,
  test.data,
  architecture,
  noise.weight = 0.5,
  brightness = 0,
  contrast = 0,
  saturation = 0,
  unfreeze.param = TRUE,
  batch_size = 32,
  learning_rate,
  save.model = FALSE,
  epoch.iterations = 1,
  early.stop = "yes",
  output.base.path = "data/",
  trainingfolder,
  list.thresholds = seq(0.1, 1, 0.1),
  positive.class = "Gibbons",
  negative.class = "Noise"
)

Arguments

input.data.path: Character. The path to the folder containing the training data.
test.data: Character. The path to the folder containing the test data.
architecture: Character. The CNN architecture to use ('alexnet', 'vgg16', 'vgg19', 'resnet18', 'resnet50', or 'resnet152').
noise.weight: Numeric. Assigned weight for the noise class. Default is 0.5.
brightness: Numeric. Brightness adjustment factor for color jitter. A value of 0 means no change. Higher values increase brightness. Default is 0.
contrast: Numeric. Contrast adjustment factor for color jitter. A value of 0 means no change. Higher values increase contrast. Default is 0.
saturation: Numeric. Saturation adjustment factor for color jitter. A value of 0 means no change. Higher values increase color saturation. Default is 0.
unfreeze.param: Logical. Determines whether to unfreeze all layers of the pretrained CNN for retraining. Default is TRUE.
batch_size: Numeric. Batch size for training the model. Default is 32.
learning_rate: Numeric. The learning rate for training the model.
save.model: Logical. Whether to save the trained model for future use. Default is FALSE.
epoch.iterations: Numeric. The number of epochs for training the model. Default is 1.
early.stop: Character. Determines whether early stopping should be applied or not. Options: "yes" or "no". Default is 'yes'.
output.base.path: Character. The base path where the output files should be saved. Default is 'data/'.
trainingfolder: Character. A descriptor of the training data used for naming output files.
list.thresholds: Numerical list indicating thresholds. Default is seq(0.1,1,.1).
positive.class: Character. The name of the positive class label. Default is 'Gibbons'.
negative.class: Character. The name of the negative class label. Default is 'Noise'.

Value

The function generates multiple output files, including:

Trained Models: If save.model = TRUE outputs saved model files (.pt) for specified architectures.
Training Logs: logs_model.csv containing logs of training sessions, including loss and accuracy metrics.
Metadata: model_metadata.csv contains metadata from training run
Model predictions: Saved for each architecture in output_TrainedModel_testdata.csv
Performance Evaluation: Saves .csv summarizing performance for each architecture in nested folder in output.base.path 'performance_tables'.

Note

Requires train, valid, and test folders created using created using 'spectrogram_images'

Examples

{
  input.data.path <- system.file("extdata", "binary/", package = "gibbonNetR")
  test.data <- system.file("extdata", "binary/test/", package = "gibbonNetR")
  result <- train_CNN_binary(
    input.data.path = input.data.path,
    test.data = test.data,
    architecture = "alexnet", # Choose architecture
    unfreeze.param = TRUE,
    batch_size = 6,
    learning_rate = 0.001,
    epoch.iterations = 1, # Or any other list of integer epochs
    early.stop = "yes",
    output.base.path = paste(tempdir(), "/", sep = ""),
    trainingfolder = "test_binary"
  )
  print(result)
}
#> Training alexnet
#> Postive class = Gibbons  and Negative class = Noise
#> Warning: Some torch operators might not yet be implemented for the MPS device. A
#> temporary fix is to set the `PYTORCH_ENABLE_MPS_FALLBACK=1` to use the CPU as a
#> fall back for those operators:
#> ℹ Add `PYTORCH_ENABLE_MPS_FALLBACK=1` to your `.Renviron` file, for example use
#>   `usethis::edit_r_environ()`.
#> ✖ Using `Sys.setenv()` doesn't work because the env var must be set before R
#>   starts.
#> Epoch 1/1
#> Train metrics: Loss: 0.292 - Acc: 0.8333
#> Valid metrics: Loss: 30.739 - Acc: 0.4444
#> `geom_line()`: Each group consists of only one observation.
#> ℹ Do you need to adjust the group aesthetic?
#> Warning: Some torch operators might not yet be implemented for the MPS device. A
#> temporary fix is to set the `PYTORCH_ENABLE_MPS_FALLBACK=1` to use the CPU as a
#> fall back for those operators:
#> ℹ Add `PYTORCH_ENABLE_MPS_FALLBACK=1` to your `.Renviron` file, for example use
#>   `usethis::edit_r_environ()`.
#> ✖ Using `Sys.setenv()` doesn't work because the env var must be set before R
#>   starts.
#> Here are actual class labels, if they do not contain the positive or negative class cannot evaluate model performance: GibbonsHere are actual class labels, if they do not contain the positive or negative class cannot evaluate model performance: Noise

#> NULL

Arguments

Value

Note

See also

Examples