0

I would like to use my tailor-made Azure Cognitive Services Custom Vision V3.0 API and/or the exported Tensorflow model to implement a find-similar-image function. Is this technically possible at the present time?

Put another way: Can I access the feature vectors/fingerprints internal to Custom Vision by any means? Can I hack the Tensorflow model to get these out?

jtlz2
  • 7,700
  • 9
  • 64
  • 114

1 Answers1

0

There is a solution call Custom Visual Search which may be able to help or I would also recommend Custom Vision. Depending on the granularity of detail in what the images need to identify.You can run the Custom Vision export version of the trained models in containers/on-prem. Please refer the following reference of the export model python.

Ram
  • 2,459
  • 1
  • 7
  • 14