Unknown people benchmarking with lfw dataset. #171

Merged
merged 7 commits into from Jul 28, 2016

Projects

None yet

2 participants

@Vijayenthiran
Contributor

What does this PR do?

Improved the previous PR on unknown classification benchmark.

Where should the reviewer start?

./evaluation/lfw-classification-unknown.py

How should this PR be tested?

  1. Preprocess the lfw dataset:

    ./evaluation/lfw-classification-unknown.py --verbose preprocess --lfwDir <lfw-directory-location> --rangeOfPeople <range-of-people-to-be-taken-as-known-dataset> --featuresDir <directory-where-features-should-be-stored>

    Eg:
    ./evaluation/lfw-classification-unknown.py --verbose preprocess --lfwDir /Users/Vijay/scikit_learn_data/lfw_home/lfw_funneled --rangeOfPeople 0:10 --featuresDir /Users/Vijay/Documents/ImageProcessing/FaceRecognition/openface/data/lfw_benchmark

  2. Train the dataset and create different classifiers:

    ./evaluation/lfw-classification-unknown.py train <feature-directory>

    Eg:
    ./evaluation/lfw-classification-unknown.py train ./data/lfw_benchmark/train_known_features

  3. Run the classification on known dataset:

    ./evaluation/lfw-classification-unknown.py inferFromTest <feature-directory> <known-test-dataset>

    is generated from the preprocess step.

    Eg:
    ./evaluation/lfw-classification-unknown.py inferFromTest ./data/lfw_benchmark/train_known_features ./data/lfw_benchmark/test_known_raw

  4. Run the classification on unknown dataset:

    ./evaluation/lfw-classification-unknown.py inferFromTest <feature-directory> <unknown-test-dataset> --threshold <threshold-value-in-float> --unknown

    is generated from preprocess step.

    Eg:
    ./evaluation/lfw-classification-unknown.py inferFromTest ./data/lfw_benchmark/train_known_features ./data/lfw_benchmark/test_unknown_raw --threshold 0.9 --unknown

Any background context you want to provide?

The PR helps to create a benchmark of unknown people classification on lfw dataset.

What are the relevant issues?

#144

Screenshots (if appropriate)

Questions:

  • Do the docs need to be updated?

    Yes, As per your previous comment
  • Does this PR add new (Python) dependencies?

    Yes. Updated in requirement.txt
@bamos bamos merged commit fa2a3a5 into cmusatyalab:master Jul 28, 2016

1 check passed

continuous-integration/travis-ci/pr The Travis CI build passed
Details
@bamos bamos referenced this pull request Jul 28, 2016
Open

Unknown people detection #144

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment