HyperAI
HomeNewsLatest PapersTutorialsDatasetsEventsWikiSOTALLM Models
Search
About
HyperAI

3D Human Pose Estimation On 3Dpw

Metrics

Acceleration Error
MPJPE
MPVPE
PA-MPJPE

Results

Performance results of various models on this benchmark

Comparison Table
Model NameAcceleration ErrorMPJPEMPVPEPA-MPJPE
live-stream-temporally-embedded-3d-human-body11.484.6100.352.3
beyond-static-features-for-temporally795111.555.8
self-attentive-3d-human-pose-and-shape77.985.8100.650.4
learning-to-regress-bodies-from-images-using-1-85.799.551.7
predicting-camera-viewpoint-improves-cross-89.7-65.2
xformer-fast-and-accurate-monocular-3d-body-7587.145.7
temporal-aware-refinement-for-video-based7.762.774.440.6
back-to-optimization-diffusion-based-zero-69.7-40.3
pose2pose-3d-positional-pose-guided-3d---54.4
learning-3d-human-pose-from-structure-and---92.2
wham-reconstructing-world-grounded-humans-57.868.735.9
deciwatch-a-simple-baseline-for-10x-efficient-75.5-46.4
3d-human-mesh-estimation-from-virtual-markers-1-67.577.941.3
learning-to-reconstruct-3d-human-pose-and-96.9116.459.2
a-lightweight-graph-transformer-network-for-88.5106.258.9
capturing-the-motion-of-every-joint-3d-human16.575.687.942
3dhr-co-a-collaborative-test-time-refinement-63.72-42.11
probabilistic-modeling-for-human-mesh---65
global-to-local-modeling-for-video-based-3d6.680.796.350.6
back-to-optimization-diffusion-based-zero-80.9-42.6
out-of-domain-human-mesh-reconstruction-via-65.58240.4
pc-hmr-pose-calibration-for-3d-human-mesh-87.8108.666.9
hulk-a-universal-knowledge-translator-for-6779.839.9
capturing-humans-in-motion-temporal-attentive7.484.399.752.1
not-all-tokens-are-equal-human-centric-visual-80.6-49.3
arts-semi-analytical-regressor-using6.567.781.446.5
ddt-a-diffusion-driven-transformer-based6.685.9101.253.3
spec-seeing-people-in-the-wild-with-an---53.2
probabilistic-3d-human-shape-and-pose-90.9-61
mesh-graphormer-74.787.745.6
3dcrowdnet-2d-human-pose-guided3d-crowd-human-85.8108.555.8
pliks-a-pseudo-linear-inverse-kinematic-60.573.338.5
mug-multi-human-graph-network-for-3d-mesh-87106.260.5
pymaf-x-towards-well-aligned-full-body-model-74.287.045.3
occluded-human-mesh-recovery-89.7107.158.3
sefd-learning-to-distill-complex-pose-and-77.3792.6049.39
non-local-latent-relation-distillation-for-1---72.1
thundr-transformer-based-3d-human-74.8-51.5
spatio-temporal-tendency-reasoning-for-human6.985.2101.252.4
accurate-3d-body-shape-regression-using-1-95.2-62.6
kinematic-aware-hierarchical-attention6.877.1--
learning-unorthogonalized-matrices-for-67.679.242.0
pose2mesh-graph-convolutional-network-for-3d22.688.9106.358.3
3d-human-pose-estimation-using-spatio-1---71.8
mpt-mesh-pre-training-with-transformers-for-65.979.442.8
weakly-supervised-3d-human-pose-and-shape-80.2-49.8
cross-attention-of-disentangled-modalities-73.584.144.6
beyond-static-features-for-temporally7.186.5102.952.7
3d-human-motion-estimation-via-motion11.686.9-54.7
multi-hmr-multi-person-whole-body-human-mesh-61.475.941.7
smpler-x-scaling-up-expressive-human-pose-and-75.2--
ivt-an-end-to-end-instance-guided-video---46
multi-person-3d-pose-and-shape-estimation-via-66.076.339.0
zolly-zoom-focal-length-correctly-for-6576.339.8
unihpe-towards-unified-human-pose-estimation---65.7
posenet3d-unsupervised-3d-human-shape-and---63.2
3d-human-reconstruction-in-the-wild-with-1-65.276.841.9
ponet-robust-3d-human-pose-estimation-via---76.2
learning-local-recurrent-models-for-human15.681.793.651.2
bilevel-online-adaptation-for-out-of-domain-77.291.249.5
motionbert-unified-pretraining-for-human-68.879.440.6
multi-initialization-optimization-network-for-81.98-52.34
sampling-is-matter-point-guided-3d-human-mesh-1-73.985.544.9
humaniflow-ancestor-conditioned-normalising-83.9-53.4
beyond-weak-perspective-for-monocular-3d-83.2-59.7
ikol-inverse-kinematics-optimization-layer-71.184.144.5
neural-descent-for-visual-3d-human-pose-and-81.4-57.5
occluded-human-body-capture-with-self-83.7110.151.7
leveraging-mocap-data-for-human-mesh-recovery8.389.4103.852.9
bopr-body-aware-part-regressor-for-human-65.480.842.5
hulk-a-universal-knowledge-translator-for-66.377.438.5
tape-temporal-attention-based-probabilistic8.979.998.151.5
3d-human-reconstruction-in-the-wild-with-1-87.3102.152.7
hybrik-x-hybrid-analytical-neural-inverse-71.682.341.8
probabilistic-modeling-for-human-mesh---55.1
cyclic-test-time-adaptation-on-monocular-64.776.739.9
end-to-end-human-pose-and-mesh-reconstruction-77.188.247.9
remocap-disentangled-representation-learning-72.781.944.1
deformable-mesh-transformer-for-3d-human-mesh-72.982.644.3
genhmr-generative-human-mesh-recovery-68.177.542.1
cliff-carrying-location-information-in-full-6981.243
kama-3d-keypoint-aware-body-mesh-articulation--97.051.1
w-hmr-human-mesh-recovery-in-world-space-with-64.675.740.5
synthetic-training-for-accurate-3d-human-pose---66.8
dual-networks-based-3d-multi-person-pose---61.7
one-stage-3d-whole-body-mesh-recovery-with-74.7-45.1
on-self-contact-and-human-pose- 84.9-55.5
hybridcap-inertia-aid-monocular-capture-of-72.1--
hybrik-a-hybrid-analytical-neural-inverse-74.186.545.0
frankmocap-a-monocular-3d-whole-body-pose-94.3-60
postometro-pose-token-enhanced-mesh-68.478.040.8
pare-part-attention-regressor-for-3d-human-74.588.646.5
smpler-taming-transformers-for-monocular-3d-73.78243.4
vibe-video-inference-for-human-body-pose-and23.482.999.151.9
encoder-decoder-with-multi-level-attention17.679.192.645.7
postometro-pose-token-enhanced-mesh-67.776.839.8
unihpe-towards-unified-human-pose-estimation---51.6
3d-human-shape-and-pose-from-a-single-low-96.36-58.98
psvt-end-to-end-multi-person-3d-pose-and-73.18443.5
keep-it-smpl-automatic-estimation-of-3d-human---106.8
potter-pooling-attention-transformer-for-7587.444.8
biopose-biomechanically-accurate-3d-pose-69.079.839.5
probabilistic-modeling-for-human-mesh---59.9
graph-and-temporal-convolutional-networks-for---64.2
learning-3d-human-shape-and-pose-from-dense-85.5-54.8
monocular-expressive-body-regression-through-93.4-60.7
end-to-end-recovery-of-human-shape-and-pose37.4130.0--
hierarchical-kinematic-probability-84.9-53.6
lasor-learning-accurate-3d-human-pose-and---57.9
motionbert-unified-pretraining-for-human-76.988.147.2
multi-view-matching-mvm-facilitating-multi---78.2
trace-5d-temporal-regression-of-avatars-with-1-79.197.337.8
staf-3d-human-mesh-recovery-from-video-with8.280.695.348.0
a-simple-yet-effective-baseline-for-3d-human---157.0
meev-body-mesh-estimation-on-egocentric-video-81.74--
cyclic-test-time-adaptation-on-monocular-84.499.951.1
poseaug-a-differentiable-pose-augmentation---73.2
instance-aware-contrastive-learning-for-73.280.344.3
exploiting-temporal-context-for-3d-human-pose---72.2
kinematic-aware-hierarchical-attention874.6--
metrabs-metric-scale-truncation-robust-68.8-49.7
implicit-3d-human-mesh-recovery-using-1-74.387.145.4
i2l-meshnet-image-to-lixel-prediction-network-130.993.2110.158.6
hybrik-transformer-71.683.642.3
deep-two-stream-video-inference-for-human1176.793.550.3
tokenhmr-advancing-human-mesh-recovery-with-a-7184.644.3
posetriplet-co-evolving-3d-human-pose-115-69.5
heater-an-efficient-and-unified-network-for-73.486.945.9
sefd-learning-to-distill-complex-pose-and-64.7578.3643.79
decomposed-human-motion-prior-for-video-pose17.189.4112.651.4
heuristic-weakly-supervised-3d-human-pose---66.1
centerhmr-a-bottom-up-single-shot-method-for-76.793.447.3
humans-in-4d-reconstructing-and-tracking-69.882.244.4
learning-3d-human-dynamics-from-video15.2116.5-72.6
exemplar-fine-tuning-for-3d-human-pose---51.6
blanketgen-a-synthetic-blanket-occlusion---53.96
inference-stage-optimization-for-cross---75.8
thundr-transformer-based-3d-human-86.8-59.9
human-mesh-recovery-from-monocular-images-via---69.5
body-meshes-as-points-104.1119.363.8
niki-neural-inverse-kinematics-with-71.386.640.6
3d-human-pose-and-shape-regression-with-92.8110.158.9