pyflink.multimodal.utils.normalize_batch

normalize_batch(batch, mean=(0.485, 0.456, 0.406), std=(0.229, 0.224, 0.225))

Normalize a batch of images.

Converts uint8 [0, 255] images to float32 and normalizes with the given mean and std. Defaults to ImageNet statistics.

Parameters

• batch – np.ndarray with shape (N, H, W, C), dtype uint8.

• mean – Per-channel means (scaled to [0, 1]). Length must match the number of channels in batch.

• std – Per-channel standard deviations (scaled to [0, 1]). Length must match the number of channels in batch.

Returns

np.ndarray with shape (N, H, W, C), dtype float32.

Example:

batch = resize_batch(pixel_arrays, (224, 224))
normed = normalize_batch(batch)  # ImageNet normalization