A small vision model processes a batch of RGB images. The batch arrives as a 4-D tensor and moves through this pipeline:
| Stage | Shape |
|---|---|
| Input batch | |
| After flatten | |
| After dense layer |
The axes of the input tensor are (batch, colour channels, height, width). The flatten step unrolls each image into one long vector, leaving the batch axis untouched. The dense layer then multiplies that flattened matrix by a weight matrix of shape , with no bias term.
How many individual numbers does contain?
Select all that apply.