pub fn validate_answer_feedback_shape(
mode: ClassifierTrainingMode,
feedback: [u32; 3],
reference: [u32; 3],
output_shapes: &[[u32; 3]],
class_count: u32,
) -> Result<(), String>Expand description
Answer feedback must match the classifier output.
Kernel mode compares the class area. Scanner mode compares each detection
twin (W×H×1, class as potential) or accepts a whole-image class of
class_count channels. With no twin yet, the mask is that output.