pose文件中的每一个pose都是啥呢？有没有说明文档？

[ciaoyizhen]

想要找一个动的稍微少一点的，适合男性的，不知道用哪个？
有没有对每个pose的说明文档？

[mengrang]

可以自行在推理代码中加入可视化

[ciaoyizhen]

看了一下，开头和结尾是连续的，是不是可以直接将pose复制多份，然后就可以生成较长的视频了呢？

[mengrang]

可以的，我们我们近期将出一个demo，可以直接从视频提取pose；也会增加一些趣味手势

[heguyu]

{'bodies': {'candidate': array([[0.49360099, 0.22189483],
[0.50113812, 0.49804543],
[0.29951984, 0.50841286],
[0.15254575, 0.82886065],
[0.37677544, 0.89106522],
[0.70275641, 0.487678 ],
[0.85726763, 0.80058584],
[0.64434364, 0.87410033],
[0.36923831, 1.07013898],
[0.24675991, 0.88729524],
[0.41257682, 0.52537774],
[0.63869079, 1.06636901],
[0.82900338, 0.8929502 ],
[0.73667351, 0.51972278],
[0.45026248, 0.18231011],
[0.53882379, 0.18419509],
[0.39938684, 0.22566481],
[0.60100513, 0.23131977]]), 'subset': array([[ 0., 1., 2., 3., 4., 5., 6., 7., 8., -1., -1., 11., -1.,
-1., 14., 15., 16., 17.]]), 'score': array([[1.04649591, 1. , 1.04879224, 1.04137862, 0.95741969,
1.01655877, 1.06691086, 0.91304374, 0.82183951, 0.12016864,
0.05385346, 0.77912098, 0.10244673, 0.09161397, 1.03984749,
1.03619802, 1.04152858, 1.01903856]])}, 'hands': array([[[0.64622792, 0.8797553 ],
[0.59912085, 0.86279041],
[0.55389805, 0.84394054],
[0.51621239, 0.85148049],
[0.48606386, 0.85525046],
[0.5482452 , 0.87598532],
[0.50490669, 0.88541026],
[0.47664244, 0.88541026],
[0.45591533, 0.88729524],
[0.55578233, 0.90991509],
[0.51055954, 0.91557005],
[0.47852673, 0.91557005],
[0.45591533, 0.91745504],
[0.56143518, 0.93441992],
[0.51998096, 0.94195987],
[0.48983243, 0.94195987],
[0.46910531, 0.94007488],
[0.56708803, 0.9513848 ],
[0.5369395 , 0.95703976],
[0.51432811, 0.95703976],
[0.49548528, 0.95326979]],

   [[0.37300688, 0.89672018],
    [0.41822967, 0.8797553 ],
    [0.4596839 , 0.86090543],
    [0.49360099, 0.86656039],
    [0.52374952, 0.86279041],
    [0.47475816, 0.88352527],
    [0.51809667, 0.89106522],
    [0.5482452 , 0.89672018],
    [0.57462517, 0.90237514],
    [0.46910531, 0.91368506],
    [0.51432811, 0.92687997],
    [0.55012949, 0.93253493],
    [0.58027801, 0.93441992],
    [0.4596839 , 0.94007488],
    [0.50302241, 0.9513848 ],
    [0.54070807, 0.95703976],
    [0.5708566 , 0.95703976],
    [0.45403105, 0.95892475],
    [0.48794814, 0.9664647 ],
    [0.51244382, 0.97023467],
    [0.54070807, 0.97023467]]]), 'hands_score': array([[1.01784444, 1.01205337, 1.00903678, 1.07593071, 0.92000687,
    0.96131736, 0.99163735, 0.9397071 , 0.8542974 , 0.94539887,
    0.98131287, 0.92457664, 0.85728598, 0.96360236, 0.96975034,
    0.91211516, 0.86370176, 0.95182276, 0.95610178, 0.96575958,
    0.87698764],
   [1.08709133, 0.99899018, 1.01771402, 1.0406872 , 0.88203788,
    1.02121532, 1.01727617, 0.94390529, 0.92001981, 1.00247347,
    1.01739991, 1.02411962, 1.04201746, 1.05057764, 1.02428138,
    1.02243328, 1.05544412, 1.05996239, 0.98840696, 0.97266912,
    0.89904326]]), 'faces': array([[[0.40127112, 0.18985006],
    [0.40127112, 0.21812486],
    [0.40503969, 0.24262969],
    [0.41069254, 0.26713452],
    [0.41822967, 0.28975437],
    [0.43330393, 0.31048922],
    [0.4483782 , 0.32745411],
    [0.4709896 , 0.33687904],
    [0.49548528, 0.34064901],
    [0.51998096, 0.33687904],
    [0.54259235, 0.32745411],
    [0.55766662, 0.31237421],
    [0.57462517, 0.29352434],
    [0.58404658, 0.27278948],
    [0.58969943, 0.24828465],
    [0.59535228, 0.22377982],
    [0.59723656, 0.19739   ],
    [0.41822967, 0.16534522],
    [0.43141965, 0.15780528],
    [0.44649392, 0.1540353 ],
    [0.46156818, 0.15592029],
    [0.47664244, 0.15969026],
    [0.51432811, 0.15969026],
    [0.52940237, 0.15592029],
    [0.54636092, 0.15592029],
    [0.56143518, 0.15969026],
    [0.57650945, 0.1691152 ],
    [0.49548528, 0.18608008],
    [0.49360099, 0.19927499],
    [0.49360099, 0.21435489],
    [0.49360099, 0.22943478],
    [0.47475816, 0.24451468],
    [0.48417958, 0.24639967],
    [0.49360099, 0.24828465],
    [0.50302241, 0.24639967],
    [0.51432811, 0.24639967],
    [0.43141965, 0.18608008],
    [0.44460963, 0.17665515],
    [0.4596839 , 0.17854013],
    [0.4709896 , 0.18985006],
    [0.45779961, 0.19362003],
    [0.44460963, 0.19173504],
    [0.51998096, 0.18985006],
    [0.53128665, 0.18042512],
    [0.54636092, 0.18042512],
    [0.5595509 , 0.18796507],
    [0.54636092, 0.19362003],
    [0.53317094, 0.19362003],
    [0.45779961, 0.27655946],
    [0.4709896 , 0.26713452],
    [0.48606386, 0.26336455],
    [0.49360099, 0.26336455],
    [0.50113812, 0.26336455],
    [0.51621239, 0.26901951],
    [0.53128665, 0.27844444],
    [0.51998096, 0.28786938],
    [0.50679097, 0.29540933],
    [0.49171671, 0.29729431],
    [0.47852673, 0.29352434],
    [0.46722103, 0.28598439],
    [0.46156818, 0.27655946],
    [0.47664244, 0.27278948],
    [0.49360099, 0.27278948],
    [0.51055954, 0.27278948],
    [0.52751809, 0.27844444],
    [0.51055954, 0.28221442],
    [0.49360099, 0.2840994 ],
    [0.47664244, 0.28032943]]]), 'faces_score': array([[1.00815272, 1.02989352, 1.02240264, 1.0185473 , 1.00841594,
    0.95284081, 1.0058614 , 0.98970169, 0.99450189, 0.9747259 ,
    1.0070163 , 0.98279119, 1.03588855, 1.02679145, 1.00932705,
    1.00541174, 1.00491977, 1.02311814, 1.02428389, 1.00662935,
    1.01114213, 1.0193795 , 0.9921959 , 1.00226426, 1.01529574,
    1.04527342, 1.02984381, 1.03507435, 1.06036317, 1.03340232,
    1.01070309, 0.99256474, 1.01235306, 1.00383842, 0.98816001,
    0.97862232, 1.01327407, 0.98359644, 0.99704206, 1.03998542,
    1.02909076, 1.02111089, 1.0268563 , 0.98998958, 1.00104976,
    1.04736447, 1.04739892, 1.03581643, 0.99167144, 1.00561106,
    1.01832843, 1.01620328, 1.01102448, 0.99719793, 1.00615883,
    0.99951684, 0.96875185, 0.97714323, 0.99151081, 1.01753688,
    0.99534529, 1.00244009, 0.9971202 , 0.98611832, 0.99992156,
    0.99956077, 1.01426053, 1.01147926]]), 'num': 1, 'draw_pose_params': [768, 768, 0, 768, 0, 768]}
    
    
    是二维坐标点，每一个npy类似一帧

转成图片就是手部点

[heguyu]

可以的，我们我们近期将出一个demo，可以直接从视频提取pose；也会增加一些趣味手势

能不能推荐一个提前手部姿势直接转成npy格式的项目，谢谢啦

[mengrang]

https://github.com/antgroup/echomimic_v2/blob/main/demo.ipynb