Conversation
algorithms/FaceApi/FaceIdentifier.cs
Outdated
| public async Task<bool> Evaluate(string groupId, double matchThreshold, string imagePath1, string imagePath2) | ||
| { | ||
| var faces = await Task.WhenAll(new[] { DetectFaces(imagePath1), DetectFaces(imagePath2) }); | ||
| var result = await Client.Face.VerifyFaceToFaceAsync(faces[0][0], faces[1][0]); |
There was a problem hiding this comment.
Would it be possible to change this so that it checks all faces in both images and if any pair is a match returns true? That way for testing we can use images that have multiple people in them.
There was a problem hiding this comment.
I thought this would be handled during the alignment step prior to running the evaluation? I just pushed an evaluation function that computes accuracy/precision/recall given a pairs.txt file. Check it out: 7d5d577.
There was a problem hiding this comment.
This might still be useful though for the custom data-set (where we don't necessarily know exactly who are all the people in the image), so I've submitted an issue for this #38.
Foxman13
left a comment
Foxman13
left a comment
There was a problem hiding this comment.
Usually, when building a CLI, there is parsing for the args that has error handling built in for missing or wrong arguments. I can see in this implementation where if no args or the wrong args are passed in, the application would crash due to NullReferenceException.
Should we add this type of pattern or are we even using the CLI?
4 participants
See #6