search

visionjo / facerec-bias-bfw

Source code and notebooks to reproduce experiments and benchmarks on Bias Faces in the Wild (BFW).
BSD 3-Clause "New" or "Revised" License
45 stars 9 forks source link

add function for overlapped score distribution plots #28

Closed suchanv closed 4 years ago

visionjo commented 4 years ago

What, at first, appeared to be an outlier, winded up being a true negative score_distribution

print(df.sort_values(by=['label', 'score'], ascending=[True, False])[['p1', 'p1','label','score']].head().to_string(index=False))
                p1                                 p1              label   score
indian_males/n008771/0158_02.jpg  indian_males/n008771/0158_02.jpg   0    0.804909
indian_males/n008311/0213_01.jpg  indian_males/n008311/0213_01.jpg   0    0.591442
indian_males/n008311/0213_01.jpg  indian_males/n008311/0213_01.jpg   0    0.583275
indian_males/n008313/0553_01.jpg  indian_males/n008313/0553_01.jpg   0    0.550018
indian_males/n008311/0524_01.jpg  indian_males/n008311/0524_01.jpg   0    0.538404

i.e., outlier indian_males/n008771/0158_02.jpg indian_males/n007193/0409_02.jpg -- 0.804909 --> true negative