Forum Discussion
Image recognition
New to AI. Is it possible to add image and OCR recognition to the search to allow images identification and comparisons
liamca-msft
Microsoft
MicrosoftHi Frederick,
Great question! There are multiple ways you can look at this and let me outline a few and you can hopefully help me drill down into what you are interested in.
First with OCR, you can absolutely use this as an option and we have this built into Azure Cognitive Search (https://docs.microsoft.com/en-us/azure/search/cognitive-search-skill-ocr) where you can extract text from an image that then gets indexed. From this point, not only can you search the content, but you can also do things like find content that is textually similar. Although there are many ways you could do this, one of the easiest is to use is the moreLikeThis api call which will get Cognitive Search to find other similar documents (https://docs.microsoft.com/en-us/azure/search/search-more-like-this).
As for other image similarities, sometimes when you are looking to compare images, it is not based on text but visual images. For example find pictures that look visually similar. For this, an easy way to approach this is to use the Vision Skill (https://docs.microsoft.com/en-us/azure/search/cognitive-search-skill-image-analysis) which allows an image to be analyzed and to extract interesting textual aspects such as "mountains" or "apple" which can be indexed into Cognitive Search. From this point you can use the search service to find other images that have similar aspects (such as find images that have mountains in them).
I hope that helps as a start, but please let me know if you are looking for anything more specific.
Liam
Great question! There are multiple ways you can look at this and let me outline a few and you can hopefully help me drill down into what you are interested in.
First with OCR, you can absolutely use this as an option and we have this built into Azure Cognitive Search (https://docs.microsoft.com/en-us/azure/search/cognitive-search-skill-ocr) where you can extract text from an image that then gets indexed. From this point, not only can you search the content, but you can also do things like find content that is textually similar. Although there are many ways you could do this, one of the easiest is to use is the moreLikeThis api call which will get Cognitive Search to find other similar documents (https://docs.microsoft.com/en-us/azure/search/search-more-like-this).
As for other image similarities, sometimes when you are looking to compare images, it is not based on text but visual images. For example find pictures that look visually similar. For this, an easy way to approach this is to use the Vision Skill (https://docs.microsoft.com/en-us/azure/search/cognitive-search-skill-image-analysis) which allows an image to be analyzed and to extract interesting textual aspects such as "mountains" or "apple" which can be indexed into Cognitive Search. From this point you can use the search service to find other images that have similar aspects (such as find images that have mountains in them).
I hope that helps as a start, but please let me know if you are looking for anything more specific.
Liam
- This is a great start. I'll take give it a look
Thanks
