deepmotionediting / deep-motion-editing Goto Github PK
View Code? Open in Web Editor NEWAn end-to-end library for editing and rendering motion of 3D characters with deep learning [SIGGRAPH 2020]
License: BSD 2-Clause "Simplified" License
An end-to-end library for editing and rendering motion of 3D characters with deep learning [SIGGRAPH 2020]
License: BSD 2-Clause "Simplified" License
Hi, when I read the source code, I find that the lambda set in the option_parser.py is different from the paper set, like the lambda_ee in the paper is 2 but is 100 in the option_parse.py. Have I missed something important?
Hi, I was kind of confused about a code snippet in the instance function "to_numpy" of class"BVH_file" in retargeting/datasets/bvh_parser.py.
from line #180 to line #184
if edge:
>>>>index = []
>>>> for e in self.edges:
>>>>>>>> index.append(e[0])
>>>>rotations = rotations[:, index, :]
rotations = rotations.reshape(rotations.shape[0], -1)
According to my understanding so far, after doing:
rotations = self.anim.rotations[:, self.corps, :]
in line #174
the variable "rotation" will hold the rotation info of the simplified skeleton. Then why we still need to do the operation from line #180 to line #184?
B.T.W could you please explain the "MOTION" part of the .bvh file for me? I searched some explainations from the website, but I didn't get a satisfied one. I guessed each line of the "MOTION" part presents the rotation information of the skeleton's joints in one frame. But how to understand the order of these numbers?
Many thanks!
如题
Hi Peizhuo,
Thanks for the guidance these days, I've reproduced part of your works.()(https://github.com/crissallan/Skeleton_Aware_Networks_for_Deep_Motion_Retargeting)
However, I haven't add the GAN part to the model(since training GAN needs lots of tricks). My training pipeline are now working in a "paired" way. Which means the .bvh file of skeleton A and B should be the same during every training iteration.
I tried to train the model in an "Un-paired" way, but the loss is really hard to converge. According to the ablation study of your paper, it states that "omitting the adversarial loss improves the performance for intra-structural skeletons"
So I'd like to know that after deprecating the adv_loss, did you train the network in a "paired"(supervised) way or still in a "Un-paired" way?
Many thanks!
Hi, I want to train style transfer with my database, I tried many times, but the program always gave me an error.
The error is "RuntimeError: CUDA error: device-side assert triggered" and "index out of bounds".
So I want to ask that how do I define a YML file?
After defining a database like xia, can I leave the rest of the code unchanged?
Thank you!
I'm really looking forward to it, thanks a lot!
Your work is great, and I'm curious about the model for the bvh data. I want to render the animation of Jerry model in your presentation video
python test.py
Batch [1/4]
Error: Numpy + Intel(R) MKL: MKL_THREADING_LAYER=INTEL is incompatible with libgomp-7c85b1e2.so.1 library.
Try to import numpy first or set the threading layer accordingly. Set NPY_MKL_FORCE_INTEL to force it.
Collecting test error...
Traceback (most recent call last):
File "test.py", line 35, in
cross_error += full_batch(0)
File "/home/linux/workspace/skeletonAware/deep-motion-editing-master/retargeting/get_error.py", line 15, in full_batch
res.append(batch(char, suffix))
File "/home/linux/workspace/skeletonAware/deep-motion-editing-master/retargeting/get_error.py", line 54, in batch
err = (pos - pos_ref) * (pos - pos_ref)
ValueError: operands could not be broadcast together with shapes (108,28,3) (156,28,3)
i want to retain the retarget model use my own dataset(eg:cmu dataset).
when i use cmu dataset ,i find split_joint.py cant split the dataset into '***_m'. so my len(characters)==1 in the retargeting/datasets/init.py
then when the code run in architecture.py->def backward_d(self)->fake = self.fake_pools[i].query(self.fake_pos[2 - i]): i will have a error:IndexError: list index out of range
so what the mean about "self.fake_pos[2 - i]"?
Hi, could you please descirbe the role of std_bvh files? I noticed that they were loaded during testing as well as training. What's that for ?
Traceback (most recent call last):
File "eval_single_pair.py", line 104, in
main()
File "eval_single_pair.py", line 92, in main
new_motion = (new_motion - dataset.mean[i][j]) / dataset.var[i][j]
RuntimeError: The size of tensor a (99) must match the size of tensor b (91) at non-singleton dimension 1
Traceback (most recent call last):
File "demo.py", line 46, in
example('Aj', 'BigVegas', '01.bvh', 'intra', './examples/intra_structure')
File "demo.py", line 42, in example
height)
File "/home/wuxiaoliang/docker/newAPP/deep-motion-editing/retargeting/models/IK.py", line 57, in fix_foot_contact
anim, name, ftime = BVH.load(input_file)
File "../utils/BVH.py", line 58, in load
f = open(filename, "r")
FileNotFoundError: [Errno 2] No such file or directory: './examples/intra_structure/result.bvh'
Hello, how can I solve this problem?@ @kfiraberman
Hi there, I got a question about the function "build_joint_topology" in skeleton.py.
I could understand that this function is mapping the "edges" to "joints" thus we could write the info of "joints" into a .bvh file.
But why we need to add "xxx_vitual" joints to the skeleton?
And I was kind of confused about the role of the variable named "out_degree" defined in the line#295.
Hi, great works! I am curious about the dataset setting: are all of Mixamo animations manually designed or are they also created by some kind of motion retargeting? It raises my doubts when some characters in Mixamo, e.g., Sword Woman, do not behave naturally in certain motions. If the Mixamo animations are not designed by human artists, then why can we take them as ground truth (please feel free to correct me)? Thanks.
when we retrained model in retargeting ,we must use the our datasets paired?
When will the training code for the retargeting part chould be released? Could you give me a rough date?
Hi,
I was kind of curious that why only send the position of root joint of each frame into the model as input?
I mean after parsing the .bvh file, we could get the positions of every joint at every frame. But we just use the root joint's position as input, is it because other joints' position info is redundant?
Mant thanks!
Hi,
When I use other characters on the Mixamo dataset in retarget work, I will need a std file, but I didn't find the code to calculate the std. Could you please provide the code to test with the customer data?
Thanks!
the style decoder take the offset as input, does it mean the offset to the parent or the last moment position
Excuse me, I use my own data for training, but there is no loss value. Is it because there is a problem with my default configuration?
====characters= [['mouse'], ['strong']]
load from file ./datasets/Mixamo/mouse.npy
Window count: 4, total frame (without downsampling): 3572
load from file ./datasets/Mixamo/strong.npy
Window count: 4, total frame (without downsampling): 3560
No scalar data was found.
Probable causes:
You haven’t written any scalar data to your event files.
TensorBoard can’t find your event files.
If you’re new to using TensorBoard, and want to find out how to add data and set up your event files, check out the README and perhaps the TensorBoard tutorial.
If you think TensorBoard is configured properly, please see the section of the README devoted to missing data problems and consider filing an issue on GitHub.
Hi, I was recently attracted by your fancy project and tried to train my custom model. However, I met a problem when preprocessing the data:
The animation data obtained from the Mixamo is in a .fbx format. But based on your source code, I think you use the .bvh data as the raw input. So could you please give me some suggestions about how to convert the .fbx data into a .bvh format!
Many Thanks!
my bvh is generate from alphapose->bvh.
i know my bvh'skeletons are different ,so how can i get my bvh to your bvh'skeletons?
excuse me,Can I just train the intra type without the cross type? @kfiraberman
What is the difference between motion retargeting and motion style transfer?
Hi, all
I read your papers carefully, you mentioned that "since the domains have different dimensions, the two networks (A → B, A → B) cannot share weights, so they had to be trained separately." in the paper.
Does this mean I need to train a new model if I want to use a character with a different number of bones from the training set?
Thanks!
Hi, sorry for bothering again,
May I ask the reason that why choose to use quaternion to represent the roration instead of the Euler angle? At first I thought that it is because if we don't know the implementation order of Euler angle, we might obtain wrong rotation, but since we could know the order of Euler angle according to the .bvh file, why we still use the quaternion representation?
Many thanks!
Error when running: sh demo.sh
System:
Windows 10 build 19.09
Python 3.7.7
1)I cloned project and downloaded the Mixamo dataset and placed it in the datasets folder
2) ran sh demo.sh
Error:
loading from epoch 20000......
load succeed!
The syntax of the command is incorrect.
Traceback (most recent call last):
File "eval_single_pair.py", line 99, in
main()
File "eval_single_pair.py", line 93, in main
model.test()
File "D:\Development\deep-motion-editing\retargeting\models\base_model.py", line 96, in test
self.compute_test_result()
File "D:\Development\deep-motion-editing\retargeting\models\architecture.py", line 296, in compute_test_result
self.writer[src][i].write_raw(gt[i, ...], 'quaternion', os.path.join(new_path, '{}_gt.bvh'.format(self.id_test)))
File "D:\Development\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 91, in write_raw
return self.write(rotations, positions, order, path, frametime, root_y=root_y)
File "D:\Development\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 80, in write
return write_bvh(self.parent, offset, rotations_full, positions, self.names, frametime, order, path)
File "D:\Development\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 10, in write_bvh
file = open(path, 'w')
FileNotFoundError: [Errno 2] No such file or directory: './pretrained/results/bvh\Aj\0_gt.bvh'
Traceback (most recent call last):
File "demo.py", line 46, in
example('Aj', 'BigVegas', 'Dancing Running Man.bvh', 'intra', './examples/intra_structure')
File "demo.py", line 42, in example
height)
File "D:\Development\deep-motion-editing\retargeting\models\IK.py", line 57, in fix_foot_contact
anim, name, ftime = BVH.load(input_file)
File "../utils\BVH.py", line 58, in load
f = open(filename, "r")
FileNotFoundError: [Errno 2] No such file or directory: './examples/intra_structure\result.bvh'
Error when running python test.py
loading from epoch 20000......
load succeed!
loading from ./pretrained/models\topology1
loading from epoch 20000......
load succeed!
0%| | 0/106 [00:00<?
he syntax of the command is incorrect.
0%| | 0/106 [00:02<?
Traceback (most recent call last):
File "eval.py", line 37, in
main()
File "eval.py", line 33, in main
model.test()
File "D:\Mocap Software\deep-motion-editing\retargeting\models\base_model.py", line 96, in test
self.compute_test_result()
File "D:\Mocap Software\deep-motion-editing\retargeting\models\architecture.py", line 296, in compute_test_re
self.writer[src][i].write_raw(gt[i, ...], 'quaternion', os.path.join(new_path, '{}_gt.bvh'.format(self.id_t
File "D:\Mocap Software\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 91, in write_raw
return self.write(rotations, positions, order, path, frametime, root_y=root_y)
File "D:\Mocap Software\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 80, in write
return write_bvh(self.parent, offset, rotations_full, positions, self.names, frametime, order, path)
File "D:\Mocap Software\deep-motion-editing\retargeting\datasets\bvh_writer.py", line 10, in write_bvh
file = open(path, 'w')
FileNotFoundError: [Errno 2] No such file or directory: './pretrained/results/bvh\BigVegas\0_gt.bvh'
Collecting test error...
Traceback (most recent call last):
File "test.py", line 35, in
cross_error += full_batch(0)
File "D:\Mocap Software\deep-motion-editing\retargeting\get_error.py", line 15, in full_batch
res.append(batch(char, suffix))
File "D:\Mocap Software\deep-motion-editing\retargeting\get_error.py", line 31, in batch
files = [f for f in os.listdir(new_p) if
FileNotFoundError: [WinError 3] The system cannot find the path specified: './pretrained/results/bvh\Mousey_m'
I have been looking for a project like this. There are many excellent projects out there that output 2D points such as OpenPose but not many that do the conversion to 3D. The ones I didn't really work well. I am excited about this project since it is specifically what I am looking for.
Thanks,
Dan
Hi, I customized a TrainDataset and tried to use this dataset to train the motion retargeting model from scratch. However, after run the train.py(it even isn't been run successfully ), then when I run the demo.py or eval_single_pair.py it throw our errors like below:
**
loading from ./pretrained/models/topology0
loading from epoch 20000......
Traceback (most recent call last):
File "/home/deep-motion-editing-new/home/deep-motion-editing-new/retargeting/eval_single_pair.py", line 98, in
main()
File "/home/deep-motion-editing-new/home/deep-motion-editing-new/retargeting/eval_single_pair.py", line 78, in main
model.load(epoch=20000)
File "/home/deep-motion-editing-new/home/deep-motion-editing-new/retargeting/models/architecture.py", line 274, in load
model.load(os.path.join(self.model_save_dir, 'topology{}'.format(i)), epoch)
File "/home/deep-motion-editing-new/home/deep-motion-editing-new/retargeting/models/integrated.py", line 83, in load
map_location=self.args.cuda_device))
File "/home/hair_gans/venv/lib/python3.6/site-packages/torch/nn/modules/module.py", line 847, in load_state_dict
self.class.name, "\n\t".join(error_msgs)))
RuntimeError: Error(s) in loading state_dict for AE:
Missing key(s) in state_dict: "enc.layers.2.0.mask", "enc.layers.2.0.weight", "enc.layers.2.0.bias", "enc.layers.2.0.offset_enc.bias", "enc.layers.2.0.offset_enc.weight", "enc.layers.2.0.offset_enc.mask", "enc.layers.2.1.weight", "dec.layers.2.1.weight", "dec.layers.2.2.mask", "dec.layers.2.2.weight", "dec.layers.2.2.bias", "dec.layers.2.2.offset_enc.bias", "dec.layers.2.2.offset_enc.weight", "dec.layers.2.2.offset_enc.mask", "dec.unpools.2.weight", "dec.enc.layers.2.0.mask", "dec.enc.layers.2.0.weight", "dec.enc.layers.2.0.bias", "dec.enc.layers.2.0.offset_enc.bias", "dec.enc.layers.2.0.offset_enc.weight", "dec.enc.layers.2.0.offset_enc.mask", "dec.enc.layers.2.1.weight".
Unexpected key(s) in state_dict: "dec.layers.1.2.bias".
size mismatch for dec.layers.0.1.weight: copying a param with shape torch.Size([192, 112]) from checkpoint, the shape in current model is torch.Size([224, 224]).
size mismatch for dec.layers.0.2.mask: copying a param with shape torch.Size([96, 192, 15]) from checkpoint, the shape in current model is torch.Size([112, 224, 15]).
size mismatch for dec.layers.0.2.weight: copying a param with shape torch.Size([96, 192, 15]) from checkpoint, the shape in current model is torch.Size([112, 224, 15]).
size mismatch for dec.layers.0.2.bias: copying a param with shape torch.Size([96]) from checkpoint, the shape in current model is torch.Size([112]).
size mismatch for dec.layers.0.2.offset_enc.bias: copying a param with shape torch.Size([96]) from checkpoint, the shape in current model is torch.Size([112]).
size mismatch for dec.layers.0.2.offset_enc.weight: copying a param with shape torch.Size([96, 72]) from checkpoint, the shape in current model is torch.Size([112, 84]).
size mismatch for dec.layers.0.2.offset_enc.mask: copying a param with shape torch.Size([96, 72]) from checkpoint, the shape in current model is torch.Size([112, 84]).
size mismatch for dec.layers.1.1.weight: copying a param with shape torch.Size([184, 96]) from checkpoint, the shape in current model is torch.Size([192, 112]).
size mismatch for dec.layers.1.2.mask: copying a param with shape torch.Size([92, 184, 15]) from checkpoint, the shape in current model is torch.Size([96, 192, 15]).
size mismatch for dec.layers.1.2.weight: copying a param with shape torch.Size([92, 184, 15]) from checkpoint, the shape in current model is torch.Size([96, 192, 15]).
size mismatch for dec.layers.1.2.offset_enc.bias: copying a param with shape torch.Size([92]) from checkpoint, the shape in current model is torch.Size([96]).
size mismatch for dec.layers.1.2.offset_enc.weight: copying a param with shape torch.Size([92, 69]) from checkpoint, the shape in current model is torch.Size([96, 72]).
size mismatch for dec.layers.1.2.offset_enc.mask: copying a param with shape torch.Size([92, 69]) from checkpoint, the shape in current model is torch.Size([96, 72]).
size mismatch for dec.unpools.0.weight: copying a param with shape torch.Size([192, 112]) from checkpoint, the shape in current model is torch.Size([224, 224]).
size mismatch for dec.unpools.1.weight: copying a param with shape torch.Size([184, 96]) from checkpoint, the shape in current model is torch.Size([192, 112]).
Process finished with exit code 1
**
Even I use version control tools roll back the code, it stills raise this error!
Have you ever met this before?
Many thanks!
================================windows length is -1
Traceback (most recent call last):
File "./datasets/preprocess.py", line 72, in
write_statistics(character, './datasets/Mixamo/mean_var/')
File "./datasets/preprocess.py", line 34, in write_statistics
dataset = MotionData(new_args)
File "/home/wuxiaoliang/docker/motion_retarget/deep-motion-editing/retargeting/datasets/motion_dataset.py", line 33, in init
new_windows = self.get_windows(motions)
File "/home/wuxiaoliang/docker/motion_retarget/deep-motion-editing/retargeting/datasets/motion_dataset.py", line 106, in get_windows
return torch.cat(new_windows)
RuntimeError: There were no tensor arguments to this function (e.g., you passed an empty list of Tensors), but no fallback function is registered for schema aten::_cat. This usually means that this function requires a non-empty list of Tensors. Available functions are [CPUTensorId, CUDATensorId, QuantizedCPUTensorId, VariableTensorId]
how to solve the problem? @kfiraberman
D:\deep-motion-editing-master\retargeting>python demo.py
命令语法不正确。
loading from ./pretrained/models\topology0
loading from epoch 20000......
load succeed!
loading from ./pretrained/models\topology1
loading from epoch 20000......
load succeed!
C:\Users\anbanglee\miniconda3\envs\avatarify\lib\site-packages\torch\nn\modules\upsampling.py:129: UserWarning: nn.Upsample is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.{} is deprecated. Use nn.functional.interpolate instead.".format(self.name))
命令语法不正确。
Traceback (most recent call last):
File "eval_single_pair.py", line 99, in
main()
File "eval_single_pair.py", line 93, in main
model.test()
File "D:\deep-motion-editing-master\retargeting\models\base_model.py", line 96, in test
self.compute_test_result()
File "D:\deep-motion-editing-master\retargeting\models\architecture.py", line 296, in compute_test_result
self.writer[src][i].write_raw(gt[i, ...], 'quaternion', os.path.join(new_path, '{}_gt.bvh'.format(self.id_test)))
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 91, in write_raw
return self.write(rotations, positions, order, path, frametime, root_y=root_y)
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 80, in write
return write_bvh(self.parent, offset, rotations_full, positions, self.names, frametime, order, path)
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 10, in write_bvh
file = open(path, 'w')
FileNotFoundError: [Errno 2] No such file or directory: './pretrained/results/bvh\Aj\0_gt.bvh'
Traceback (most recent call last):
File "demo.py", line 46, in
example('Aj', 'BigVegas', 'Dancing Running Man.bvh', 'intra', './examples/intra_structure')
File "demo.py", line 42, in example
height)
File "D:\deep-motion-editing-master\retargeting\models\IK.py", line 57, in fix_foot_contact
anim, name, ftime = BVH.load(input_file)
File "../utils\BVH.py", line 58, in load
f = open(filename, "r")
FileNotFoundError: [Errno 2] No such file or directory: './examples/intra_structure\result.bvh'
(avatarify) D:\deep-motion-editing-master\retargeting>sh demo.sh
'sh' 不是内部或外部命令,也不是可运行的程序
或批处理文件。
(avatarify) D:\deep-motion-editing-master\retargeting>python test.py
Batch [1/4]
命令语法不正确。
loading from ./pretrained/models\topology0
loading from epoch 20000......
load succeed!
loading from ./pretrained/models\topology1
loading from epoch 20000......
load succeed!
0%| | 0/106 [00:00<?, ?it/s]C:\Users\anbanglee\miniconda3\envs\avatarify\lib\site-packages\torch\nn\modules\upsampling.py:129: UserWarning: nn.Upsample is deprecated. Use nn.functional.interpolate instead.
warnings.warn("nn.{} is deprecated. Use nn.functional.interpolate instead.".format(self.name))
命令语法不正确。
Traceback (most recent call last):
File "eval.py", line 37, in
main()
File "eval.py", line 33, in main
model.test()
File "D:\deep-motion-editing-master\retargeting\models\base_model.py", line 96, in test
self.compute_test_result()
File "D:\deep-motion-editing-master\retargeting\models\architecture.py", line 296, in compute_test_result
self.writer[src][i].write_raw(gt[i, ...], 'quaternion', os.path.join(new_path, '{}_gt.bvh'.format(self.id_test)))
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 91, in write_raw
return self.write(rotations, positions, order, path, frametime, root_y=root_y)
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 80, in write
return write_bvh(self.parent, offset, rotations_full, positions, self.names, frametime, order, path)
File "D:\deep-motion-editing-master\retargeting\datasets\bvh_writer.py", line 10, in write_bvh
file = open(path, 'w')
FileNotFoundError: [Errno 2] No such file or directory: './pretrained/results/bvh\BigVegas\0_gt.bvh'
Collecting test error...
Traceback (most recent call last):
File "test.py", line 35, in
cross_error += full_batch(0)
File "D:\deep-motion-editing-master\retargeting\get_error.py", line 15, in full_batch
res.append(batch(char, suffix))
File "D:\deep-motion-editing-master\retargeting\get_error.py", line 31, in batch
files = [f for f in os.listdir(new_p) if
FileNotFoundError: [WinError 3] 系统找不到指定的路径。: './pretrained/results/bvh\Mousey_m'
what a great work!
Could you please give us the training method and the retarget datasets
As seen in the demo (same animation bvh file exists to both input and destination character) with intra, is it necessary to have the same animation's bvh file for the destination character as well(Goblin_m', 'Goblin_m)or(BigVegas', 'BigVegas)?
how to slove the problem? @ @kfiraberman
Hi there, I recently met a problem during training the retargeting model by myself. That is the "ee_loss" always became "nan" type during training,
Have you ever met this problem before?
Many thansk!
From looking at both videos I see that this is being designed for "intelligent ai retargeting and stylized motion. Can the strength of the stylized motion be adjusted? Is this project related to FACS AU as far as using emotions for mocap instead of direct xyz?
Thanks!
Hello, as per the title
import sys
import os
sys.path.append("../utils")
sys.path.append("./")
import BVH
import numpy as np
import bpy
import mathutils
import pdb
#scale factor for bone length
global_scale = 10
class BVH_file:
def __init__(self, file_path):
self.anim, self.names, self.frametime = BVH.load(file_path)
#permute (x, y, z) to (z, x, y)
tmp = self.anim.offsets.copy()
self.anim.offsets[..., 0] = tmp[..., 2]
self.anim.offsets[..., 1] = tmp[..., 0]
self.anim.offsets[..., 2] = tmp[..., 1]
BVH.load(file_path) refers to import BVH, which is not in current Blender (2.83).
What python package is that?
Is it possible to train the model combining your.npy files(mixamo characters) with custom rig .npy(new custom character)?
Hi, after reading the implementation code of the skeleton.py. I got some questions which I'd like to disccuss with you.
1.Could I regard the "SkeletonConv" as a binary mask which created based on the neighboring list of each joint. If joint A is the neighbor of joint B, when convolve joint B, the binary mask on joint A is 1 else the mask is set to 0.
2.Does the "SkeletonUnpool" just duplicate the features of the pooled joint to increase the nodes of the skeleton graph?
Many thanks!
In styleTransfer,when i use other BVHs,why cant run the style_transfer/demo.sh successfully?
is my skeleton in my BVH is lack? the skeleton must match it that you give?
First of all, thanks for all the work.
Currently we can use style transfer from JSON output as given in README.md
Can we use the JSON output from OpenPose, for Motion retargeting?
So the pipeline will be, video -> json_output_from_openpose -> (bvh?) -> any_of_the_given_mixamo_models
which version of pytorch should i install
in the skining step ,i have some question.
Hello there
First of all kudos for this interesting project. I have tons of animation and from time to time I bang my head on how to port them gracefully...
I'm try to run the demo but it fails.
user:retargeting source demo.sh
Traceback (most recent call last):
File "eval_single_pair.py", line 2, in <module>
import torch
ImportError: No module named torch
Traceback (most recent call last):
File "demo.py", line 49, in <module>
example('Aj', 'BigVegas', 'Dancing Running Man.bvh', 'intra', './examples/intra_structure')
File "demo.py", line 45, in example
height)
File "/Users/max/Developer/Library/Graphics/moremocap/deep-motion-editing/retargeting/models/IK.py", line 57, in fix_foot_contact
anim, name, ftime = BVH.load(input_file)
File "../utils/BVH.py", line 58, in load
f = open(filename, "r")
FileNotFoundError: [Errno 2] No such file or directory: './examples/intra_structure/result.bvh'
Seems it cannot find pytorch, but my pip shows is installed:
user:retargeting pip list
Package Version
---------------- --------
appnope 0.1.0
backcall 0.1.0
certifi 2019.3.9
decorator 4.4.2
future 0.18.2
ipython 7.13.0
ipython-genutils 0.2.0
jedi 0.17.0
macpack 1.0.3
numpy 1.16.2
parso 0.7.0
pexpect 4.8.0
pickleshare 0.7.5
pip 20.1.1
prompt-toolkit 3.0.5
ptyprocess 0.6.0
Pygments 2.6.1
scipy 1.4.1
setuptools 39.0.1
six 1.14.0
torch 1.5.0
tqdm 4.46.1
traitlets 4.3.3
wcwidth 0.1.9
wheel 0.34.2
Anyway, if I run eval_single_pair.py directly it works:
user:retargeting python eval_single_pair.py
usage: eval_single_pair.py [-h] [--save_dir SAVE_DIR]
[--cuda_device CUDA_DEVICE]
[--num_layers NUM_LAYERS]
[--learning_rate LEARNING_RATE] [--alpha ALPHA]
[--batch_size BATCH_SIZE] [--upsampling UPSAMPLING]
[--downsampling DOWNSAMPLING]
[--batch_normalization BATCH_NORMALIZATION]
[--activation ACTIVATION] [--rotation ROTATION]
[--data_augment DATA_AUGMENT]
[--epoch_num EPOCH_NUM] [--window_size WINDOW_SIZE]
[--kernel_size KERNEL_SIZE]
[--base_channel_num BASE_CHANNEL_NUM]
[--normalization NORMALIZATION] [--verbose VERBOSE]
[--skeleton_dist SKELETON_DIST]
[--skeleton_pool SKELETON_POOL]
[--extra_conv EXTRA_CONV]
[--padding_mode PADDING_MODE] [--dataset DATASET]
[--fk_world FK_WORLD] [--patch_gan PATCH_GAN]
[--debug DEBUG] [--skeleton_info SKELETON_INFO]
[--ee_loss_fact EE_LOSS_FACT] [--pos_repr POS_REPR]
[--D_global_velo D_GLOBAL_VELO]
[--gan_mode GAN_MODE] [--pool_size POOL_SIZE]
[--is_train IS_TRAIN] [--model MODEL]
[--epoch_begin EPOCH_BEGIN]
[--lambda_rec LAMBDA_REC]
[--lambda_cycle LAMBDA_CYCLE]
[--lambda_ee LAMBDA_EE]
[--lambda_global_pose LAMBDA_GLOBAL_POSE]
[--lambda_position LAMBDA_POSITION]
[--ee_velo EE_VELO] [--ee_from_root EE_FROM_ROOT]
[--scheduler SCHEDULER]
[--rec_loss_mode REC_LOSS_MODE]
[--adaptive_ee ADAPTIVE_EE]
[--simple_operator SIMPLE_OPERATOR]
[--use_sep_ee USE_SEP_EE] [--eval_seq EVAL_SEQ]
--input_bvh INPUT_BVH --target_bvh TARGET_BVH
--test_type TEST_TYPE --output_filename
OUTPUT_FILENAME
eval_single_pair.py: error: the following arguments are required: --input_bvh, --target_bvh, --test_type, --output_filename
That means a python scripts run directly finds pytorch ok. Maybe could be something related to os.system calls? My setup:
Mac Os X 10.14.3 Mojave
default python.org Python 3.7.0, no envs
Thank you
Hi, thanks for sharing such a great project! Howver, I'd like to know that is there any way to obtain or create some customized trainig data?
Because I'd like to reproduce the fully pipeline of the project!
Many thanks!
I use reallusion iClone and Character Creator which are similar to DAZ3D.
What I am looking for is to have full body motion capture face and hands from a web camera.
One of the problems I have been running into using Character Creator 3 armature is that the facial movement is mostly based on BlendShapes (Morphs) and I don't know how to translate that from xyz coordinates. Actually I am just learning 3D and motion capture but the price for markerless mocap is high and I am just a hobbyist so I have been searching for an opensource solution.
Your software looks it will be fantastic!
Thanks,
Dan
Hi,
I have a question about the mean and var located in the ./retargeting/datasets/Mixamo/mean_var.
I'd like to know the mean and var of each character's motion data are just calculated based on the test data or they are calculated based on the whole dataset?
If it is the whole dataset, how many data did you use to calculate the mean and var?
Many thanks!
Hi=.=, it's me again,
Could you please describe the role of the function "find_seq" defined in the class "SkeletonPool" in skeleton.py as well as the role of the attribute "self.seq_list" for me?
According to my understanding so far, after the Skeleton pooling, we need a new topology to help us to calculate the neigboring list/matrix for the next Skeleton convolution layer.
But I haven't make it clear after reading the code.
Many thanks!
Hi,
I was recently reading the source code very carefully in order to understand the whole pipeline of your work.
I was kind of confused that why init a variable named "global_part_neighbor" in the in the function "find_neighbor" in the line #373 of the models/skeleton.py?
Besides, why append a "edge_num" into those global_part_neighbor's list in line #375?
Many thanks!
A declarative, efficient, and flexible JavaScript library for building user interfaces.
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
An Open Source Machine Learning Framework for Everyone
The Web framework for perfectionists with deadlines.
A PHP framework for web artisans
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
Some thing interesting about web. New door for the world.
A server is a program made to process requests and deliver data to clients.
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
Some thing interesting about visualization, use data art
Some thing interesting about game, make everyone happy.
We are working to build community through open source technology. NB: members must have two-factor auth.
Open source projects and samples from Microsoft.
Google ❤️ Open Source for everyone.
Alibaba Open Source for everyone
Data-Driven Documents codes.
China tencent open source team.