For a s3 bucket, how do you find a similar or identical image, based on image features (rather than file size)?
By features, I mean something more like SIFT than label detection. (The latter is not useful for uniqueness identification if the bucket contains a number of images primarily of the same label class, shoes, for example.)
- are there existing pipelines?
- if not, is there something more production ready than, building a solution via indexing each as a hash in a database and then comparing a new uploaded image against it?
What do you mean by “find a similar or identical image, based on image features”? Can you provide a few examples of what you mean?
Amazon Rekognition can Detect labels in an image – Amazon Rekognition, but you would end up comparing labels rather than comparing the actual images.
yeah labels don’t work since the subset could include many images with the same label but have different image features
@JohnRotenstein feature extraction that ideally can distinguish between similar but different images – perhaps more like SIFT than label extraction
There is no such capability in S3. You would need to develop your own Machine Learning model to achieve this, or find some other fancy software that might do it for you.