In SAS Viya's deep learning pipeline, the DLPy ImageTable.load_images() method loads images from a local directory into a CAS in-memory table. What additional CAS action is typically needed to convert raw loaded images to model input format?
Select an answer to reveal the explanation.
Short Explanation
Here's the deal — load_images() just pulls raw image bytes into CAS, so you still need to resize them to the model's expected height and width, and sometimes normalize the pixel values.
Full Explanation
load_images() loads raw image bytes into a CAS binary table, typically into a '_image_' binary column, but images may be different sizes and formats. The correct approach is to call the ImageProcessing.resize() action set, or use DLPy's ImageTable.resize() helper, to resize images to expected height and width dimensions, and optionally normalize pixel values. Normalization, such as scaling 0-255 to 0-1 or standardizing to mean and standard deviation, may also be needed. The option saying no additional action is needed because load_images() automatically resizes and normalizes images is wrong; it does not auto-resize to model dimensions. The option requiring PROC CONTENTS on the CAS image table to verify image bytes loaded correctly is wrong; PROC CONTENTS is not required. The option requiring a CAS session restart after loading images to initialize GPU memory allocation is wrong; a restart is not needed.