Added computation of the FID score #23
Open
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Adapted score.py to also compute the Fréchet Inception Distance using code provided by the authors of the measure.
AlexeyG
reviewed
Sep 14, 2019
| x_bin = tf.read_file(x_fp) | ||
| x_samps = tf.contrib.ffmpeg.decode_audio(x_bin, tf_ffmpeg_ext, 16000, 1)[:, 0] | ||
| x = graph.get_tensor_by_name('x:0') | ||
| activations = graph.get_tensor_by_name('classifier/dense/BiasAdd:0') |
There was a problem hiding this comment.
It looks like you are taking the logits (a 10-D vector) and compute the FID score on them. Shouldn't this be the activations of one of the earlier pooling layers (e.g. the pool4 layer) as in the original FID score?
However, one issue that would arise if an earlier layer is taken is that the dimensionality of the activations would be too high (it is already 8192 for the pool4 layer) to compute the FID score in a reasonable time. This is because complexity of the matrix square root computation is cubic in the matrix dimensionality.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Adaptation of score.py to also compute the Fréchet Inception Distance using code provided by the authors of the measure.
It makes sense to add it side to side to the evaluation of the inception measure since the two don't correlate precisely but are built on the same classifier network. In the future, it would be nice to have cleaner code and a better documentation. As it is, there is a lot of repetition between this file and score.py, but I think it's better to push it now than wait until it's perfect.