We are using VNRecognizeTextRequest to detect text in documents, and we have noticed that even in some very clear and well-formatted documents, there are still instances where text blocks are missed. the live text also have the same issue.