Revision history [back]

A CNN can do classification or regression, it depends on what you are building it for. If you wanted to look at similarity between images, you can perhaps explore one shot learning approaches. You really need to be clear about your objective here. Is it that you need to find similarity between a few / small group of Chinese words or is it detecting similarity between any word? More importantly, why are you finding similarity? If you are trying to see if one word matches another, then a classification type outcome is OK. However, if you are trying to detect if one piece of hand writing compared to another was written by the same person, then the challenge and associated solution would be very different.

Azure Cognitive Services OCR is not limited to English and does OCR Chinese characters. But you need to use the OCR service. There is a new Recognise Text service which is English only, but the OCR works fine. I have tested it at this location with the below sample image:

https://northeurope.dev.cognitive.microsoft.com/docs/services/5adf991815e1060e6355ad44/operations/56f91f2e778daf14a499e1fc/console {"url":"http://www.words-chinese.com/images2/chinese_symbols_for_power_6637_2_110.png"}

A CNN can do classification or regression, it depends on what you are building it for. If you wanted to look at similarity between images, you can perhaps explore one shot learning approaches. You really need to be clear about your objective here. Is it that you need to find similarity between a few / small group of Chinese words or is it detecting similarity between any word? More importantly, why are you finding similarity? If you are trying to see if one word matches another, then a classification type outcome is OK. However, if you are trying to detect if one piece of hand writing compared to another was written by the same person, then the challenge and associated solution would be very different.

Azure Cognitive Services OCR is not limited to English and does OCR Chinese characters. But you need to use the OCR service. There is a new Recognise Text service which is English only, but the OCR works fine. I have tested it at this location with the below sample image:

https://northeurope.dev.cognitive.microsoft.com/docs/services/5adf991815e1060e6355ad44/operations/56f91f2e778daf14a499e1fc/console https://northeurope.dev.cognitive.microsoft.com/docs/services/5adf991815e1060e6355ad44/operations/56f91f2e778daf14a499e1fc/console

{"url":"http://www.words-chinese.com/images2/chinese_symbols_for_power_6637_2_110.png"}