I'm looking for someone with experience with the Tensor Flow library for image recognition. I want to create a small specific test just to see how accurate it is (comparing 2 files: one PDF file with peoples faces and names, and another file which is a photo of the people listed in the first PDF file). I would like to know how accurate it would be, assuming that I don't have additional images of those people which I could use to 'train' the algorithm. Thanks
In my experience with AI learning models, there is one defined amount of training process to 'train' the algorithm. Some models depending on complexity will work with a little amount of training such as the one PDF file. It will greatly depend on the complexity you are looking for. Such as if you are trying to determine specific features, or tell the difference between one face or another. The model will work more effectively with more training data involved. If you are just looking for a face than this one PDF file may be sufficient. The answer is not a straight forward yes or no without further details into what type of data you are looking for as a differental identifier for each face.
Answered 3 years ago