PR #1837 expects a batch dimension or no batching at all when parsing input data. However, our OD tutorial passes a single image as [image] at the beginning, which throws plot_image_gallery.

We should remove the list wrapper and immediately pass around

image_batch = np.expand_dims(image, 0)

since there's a lot of downstream logic about image batches anyway.

See gist.

Update: I get the same issue in 0.6.1 and 0.5.1 (gist).