Comments (4)
Thanks for using our code!
About Q1, I assume you are wondering about the dimensions of theta_b^3D, theta_h^3D, ... Here referring to the orginial paper of SMPL-X:
The total number of model parameters in SMPL-X is 119: 75 for the global body rotation and { body, eyes , jaw } joints, 24 parameters for the lower dimensional hand pose PCA space, 10 for subject shape and 10 for the facial expression,
where theta_b^3D and theta_f^3D are corresponding to the global body rotation and { body, eyes , jaw } joints
, theta_h^3D is corresponding to the lower dimensional hand pose PCA space
, beta is corresponding to the subject shape
and psi is corresponding to the facial expression
. While in practical use, theta_b^3D is encoded to a lower dimension latent space by the VPoser proposed in the same paper, with size of 32. Therefore, all of them are not joint coordinates. And since SMPL-X has assembled the latent code decoding process, no conversions need to be done before feeding them to the model.
About Q2, like described in the smplify-x instruction, you could run SMPLify-X on the dataset with filtered pose as:
python smplifyx/main.py --config cfg_files/fit_smplx.yaml
--data_folder DATA_FOLDER
--output_folder OUTPUT_FOLDER
--visualize="True/False"
--model_folder MODEL_FOLDER
--vposer_ckpt VPOSER_FOLDER
--part_segm_fn smplx_parts_segm.pkl
where the DATA_FOLDER should contain two subfolders, images, where the images are located, and keypoints, where the filtered OpenPose output should be stored.
About Q3, it is possible, and there are multiple ways. The version of SMPLify-X that we used didn't save the 3D joint locations of the recovered body model. You could modify the SMPLify-X code to make it save the joint location, or you could use this function to extract the 3D joint location from the recovered obj
file, where J_regressor
is actually the files in the Step. 4 in the Installation section.
DJ-RN/script/generate_utils.py
Line 79 in 4bfdf0c
from dj-rn.
Thanks for the patient reply. I wonder if I understand it correctly: the inputs to the smpl-x model is the coordinates of the detected 2D keypoints, which will be automatically converted to the parameters of the 3D model e.g. theta_h^3D, theta_b^3D, etc. With the 3D parameters, we can get the corresponding mesh or the vertices from the smpl-x model.
from dj-rn.
The inputs to SMPLifiy-X are the coordinates of the detected 2D keypoints, and SMPLifiy-X will perform the recovery and return the parameters of the SMPL-X 3D model e.g. theta_h^3D, theta_b^3D, etc. And the SMPL-X model takes the parameters as input and output the corresponding mesh and vertices, while the joints could also be fetched with the mesh and vertices.
from dj-rn.
Got it, thank you very much
from dj-rn.
Related Issues (20)
- while the pkl file exists if not os.path.exists(os.path.join(args.res, 'results/HICO_train2015_%08d/%03d.pkl' % (key, item[6]))): continue doesn't work as expected and generate_3D_obj_GT.py creates no obj file HOT 3
- The visualized attention HOT 4
- result of generate_3D_obj_GT.py look weird HOT 1
- raise TypeError("Invalid comparison of non-real %s" % me) for line otri, obj_vertexs = get_param(result, hbox, obox, htri, img, radius, gamma_min, gamma_max) HOT 3
- Calculate ratio, gamma_min and gamma_max HOT 1
- very low GPU utilization when running the Feature_extraction.py script HOT 1
- Not sure what does the result of generate_3D_obj_Neg.py is supposed to be as or why we run this script at all? HOT 1
- File "script/concat_w2v.py", line 80, in <module> hoi_class = Trainval_Neg[key][i][1][0] TypeError: 'int' object is not subscriptable HOT 3
- --> 128 point[1] *= -1 129 projected = np.matmul(camera_transform, point) 130 img_point = projected[:2] / projected[2] IndexError: invalid index to scalar variable.
- File "/scratch3/research/code/DJ-RN/script/generate_utils.py", line 131, in point_align_vis point[1] *= -1 IndexError: invalid index to scalar variable. for Demo.ipynb HOT 3
- how to create the path including object_Neg for being able to run the concat_w2v.py script? HOT 1
- Object Sphere Center HOT 4
- Trainval_Neg.pkl file HOT 3
- What is shape and pose parameter? HOT 6
- How do I get the output of open pose for dataset ambiguous? HOT 1
- plot visualized map of a output tensor of network HOT 1
- operands could not be broadcast together with shapes (1560224,) (1489274,) HOT 2
- Do you consider human_thres and Object_thres? HOT 2
- Run SMPLify-x on vcoco dataset
- <lambda>() takes 1 positional argument but 2 were given HOT 4
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dj-rn.