For non-latin languages you have to use VNRecognizeTextRequest with accurate as the recognitionLevel.

3DBodypose is right now the only request that takes depth into account