Skip to content

Failure Cases

Viet edited this page Jun 17, 2019 · 1 revision

Failure Cases

Baseline Model

False Negative

Reasons for false negative:

  • relatively small images with a lot of black space
  • relatively little noise
  • common hands except for the arm without hand (why can the model reconstruct this so good?) and one with only parts of the hand

baseline_false_negative

False Positive

Reasons for false positive:

  • a lot of noise around the hand → masking might help
  • Maybe exclude the black image with noise. We can see that the model can distinguish between hands/arms and other things

baseline_false_positive