Details for Multi-HMR SMPL-X

AuthorsBaradel*, Fabien; Armando, Matthieu; Galaaoui, Salma; Brégier, Romain; Weinzaepfel, Philippe; Rogez, Grégory; Lucas*, Thomas
Paper titleMulti-HMR: Multi-Person Whole-Body Human Mesh Recovery in a Single Shot
Publication venuearXiv
URLhttps://github.com/naver/multi-hmr
Additional Information Single shot method (detection & regression) taking as input a RGB image at resolution 1288x1288. A ViT-L/14 is used to encode the image, then we use our HPH module for regressing human properties.

Metrics

Full setKid
F1 Score0.940F1 Score0.620
Precision0.960Precision0.470
Recall0.920Recall0.900
Body
Full setKid
MPJPE60.300MPJPE71.300
MVE59.600MVE72.000
NMJE64.100NMJE115.000
NMVE63.400NMVE116.100
Face
Full setKid
MPJPE29.200MPJPE29.200
MVE27.700MVE27.500
Full body
Full setKid
MPJPE95.700MPJPE105.300
MVE95.900MVE106.300
NMJE101.800NMJE112.000
NMVE102.000NMVE113.100
Hands
Full setKid
left hand MPJPE38.100left hand MPJPE36.000
left hand MVE40.200left hand MVE37.400
right hand MPJPE39.000right hand MPJPE36.700
right hand MVE40.900right hand MVE38.100