Apply uncertainty filtering to Bayesian CNN or webcam gaze outputs — filter_gazepoint_cnn

Provides a lightweight post-processing helper for externally generated webcam/CNN gaze predictions. The function does not train a CNN. It filters or down-weights frame-level gaze estimates using an uncertainty column.

Usage

filter_gazepoint_cnn_uncertainty(
  data,
  x,
  y,
  uncertainty = NULL,
  max_uncertainty = NULL,
  weight_output = "cnn_uncertainty_weight",
  valid_output = "cnn_valid_frame"
)

Arguments

data: A data frame containing frame-level gaze predictions.
x: Predicted x-coordinate column.
y: Predicted y-coordinate column.
uncertainty: Optional uncertainty column.
max_uncertainty: Optional maximum allowed uncertainty.
weight_output: Name of the output weight column.
valid_output: Name of the output validity column.

Value

Data frame with uncertainty weights and validity flags.