Multi-Human Parsing Track

1. Metrics

We use two human-centric metrics for multi-human parsing evaluation, which are initially reported by the MHP v1.0 paper. The two metrics are Average Precision based on part (APp) (%) and Percentage of Correctly parsed semantic Parts (PCP) (%).

2. Semantic Category Definition

  1. Background
  2. Cap/Hat
  3. Helmet
  4. Face
  5. Hair
  6. Left-arm
  7. Right-arm
  8. Left-hand
  9. Right-hand
  10. Protector
  11. Bikini/Bra
  12. Jacket/Windbreaker/Hoodie
  13. T-shirt
  14. Polo-shirt
  15. Sweater
  16. Singlet
  17. Torso-skin
  18. Pants
  19. Shorts/Swim-shorts
  20. Skirt
  1. Stockings
  2. Socks
  3. Left-boot
  4. Right-boot
  5. Left-shoe
  6. Right-shoe
  7. Left-highheel
  8. Right-highheel
  9. Left-sandal
  10. Right-sandal
  11. Left-leg
  12. Right-leg
  13. Left-foot
  14. Right-foot
  15. Coat
  16. Dress
  17. Robe
  18. Jumpsuits
  19. Other-full-body-clothes
  20. Headware
  1. Backpack
  2. Ball
  3. Bats
  4. Belt
  5. Bottle
  6. Carrybag
  7. Cases
  8. Sunglasses
  9. Eyeware
  10. Gloves
  11. Scarf
  12. Umbrella
  13. Wallet/Purse
  14. Watch
  15. Wristband
  16. Tie
  17. Other-accessaries
  18. Other-Upper-Body-Clothes
  19. Other-Lower-Body-Clothes

3. Dataset Examples