timctho / vnect-tensorflow Goto Github PK
View Code? Open in Web Editor NEWLicense: Apache License 2.0
License: Apache License 2.0
Thanks for this code. I am trying to implement this paper on the top of your code. In the paper, they mentioned that till res4f the input will be divided by 8 but when I checked the resnet50 http://ethereon.github.io/netscope/#/gist/db945b393d40bfa26006, I see that till the res4f the output will be divided by 16 ? !
I find that the documentation is lacking and haven't really been able to run it.
I've done the following
Yet I can't run any of the demos, I'm getting a dataLoss error, more specifically:
DataLossError (see above for traceback): Unable to open table file models/weights/vnect_tf: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
Is there something I need to do beforehand or did I do something wrong? I'd be willing to update the readme or even create a script to automate the process if I get this running.
Awesome solution! Do you know the minimum configuration for a pc to work with it? Do you think it would work on smartphones?
Hi,
I couldn't run the demo becouse I don't have the vnect_tf.
Readme.md say: "1.Download model, put them in folder models/weights", but can not find the link to download.
Yuri
[email protected]
Hi! I am a bit of confused while reading your admirable work that I can't catch the point of the following part in demo.py:
Lines 32 to 33 in 9aa423e
The original code meets errors when running under two scale values. And I simply try deleting either of these two values in order to fix those errors, which successfully generates relatively good results, too.
I wonder the idea of applying two kinds of scale values in this work. Could you tell me the motivation and the advantages of this strategy? Thank you very much!
Thanks you for afford TF version project. The 3D label shape formulate : (batch, W, H, nums_points), how to generate this labels?
Can anyone share weight and model files?I also have sent a emial to the author yesterday, but he haven't replied yet. my email is
[email protected]
Hello
I am Junkyung Kim.
I had tried your code.
But I got really bad estimation results with transformed weights from original parameter.
I guess the transformed weights are not exactly same as the original caffe weights.
Can somebody help me..?
I would love to run this repo on Windows. Anyone achived it?Thank you
i converted the caffemodel using the caffe_weights_to_pickle.py.
however, when i want to load the vnect.pkl with tensorflow, i get:
DataLossError (see above for traceback): Unable to open table file models/weights/vnect_tf: Data loss: not an sstable (bad magic number): perhaps your file is in a different file format and you need to use a different restore operator?
thanks for any help
@timctho As explained in the VNect official paper about Bounding Box tracker and utilizing the previously detected key points to determine the Bounding Box for the new frame. Have you implemented this idea in your code.
I am not able to understand the way you have did it. Could you please explain or refer me any article or any other resource where I could learn more about it..
Thanks in Advance
could you attach the model?
I'm having a really hard time to get caffe running on my system.
Is there a way to not use caffe?
Would it be possible to convert the model to a tensorflow model and save it's weights (through serialisation in tensorflow)?
@timctho Thanks for sharing your code. I don't see the use of Levenberg Marquardt optimization as indicated by the authors.
Hey,
I've been trying out this project and managed to get the demo working for pictures without any issues, however if I try to run the demo with a webcam instead of a picture I get weird results.
No matter what is the font or if there's a subject in the frame, I get false heatmap (garbage) all over the picture. Did anyone manage to work with a video feed from a camera or is it a known issue? I compared the code of the image and video demo and they are both pretty much identical treatment wise.
I added a picture just to show what I mean by garbage, this is not static it changes at each iteration.
hi timctho
i test VNECT-tensorflow , i find Vnect-tf very faster but less accurate . i want to know about your tensorflow model come from . how i do train myself mode .?
Hi @timctho
I had read the vnect_model.py code, and I think maybe a mistake at line 68.
self.res3d = tf.add(self.res3d_branch2c, self.res3b, name='res3d_add')
the add function second param 'self.res3b', maybe is 'self.res3c'?
How can I do this with kinect V1? I know it's for kinect alternative
Hello
Thank you for your implementation.
I am trying to run the demo, but getting errors because of the absence of model_file.
Is there anywhere that I can download the pre-trained weight file?
Thank you
It is amazing project! Thanks for sharing!
I want to ask is it good for real time pose detection?
hi timctho
/run_demo_tf.sh but i got an error
File "demo_tf.py" . line 10, in
from models.nets import vnect_model_bn_folded as vnect_model
ImportError: No module named models.nets
i download the tensorflow model abd then save models/weights/vnect_tf
my system:
ubuntu 16.04 64bit gtx1080ti cuda-8.0 python2.7 opencv 3.3.0
i don't know why .pls help me thanks
I cant use the username and password to access the download page can you send me a email [email protected]
Hello.
I have a question.
Will this library work if it only recognizes the upper body?
Config:
PS C:\Users\TechRova\Downloads\VNect-tensorflow-master> python demo_tf_gl.py --device=gpu --demo_type=webcam --test_img=
test_imgs/yuniko.jpg --model_file=models/weights/vnect_tf --plot_2d=True --plot_3d=True
pygame 1.9.6
Hello from the pygame community. https://www.pygame.org/contribute.html
Traceback (most recent call last):
File "demo_tf_gl.py", line 247, in <module>
glMatrixMode(GL_PROJECTION)
File "C:\Users\TechRova\AppData\Local\Programs\Python\Python37\lib\site-packages\OpenGL\platform\baseplatform.py", line 409, in __call__
return self( *args, **named )
File "C:\Users\TechRova\AppData\Local\Programs\Python\Python37\lib\site-packages\OpenGL\error.py", line 232, in glCheckError
baseOperation = baseOperation,
OpenGL.error.GLError: GLError(
err = 1282,
description = b'invalid operation',
baseOperation = glMatrixMode,
cArguments = (GL_PROJECTION,)
)
I can;t run this code cause there's no weights
I've sent a email to VNECT author but can;t get a reply..
So could anybody send me a email of model files??
Hi, I've successfully managed to get the code running in Ubuntu and on MacOS Mojave, I would like to integrate it into a browser application some like this: https://github.com/tensorflow/tfjs-models/tree/master/posenet any ideas how I can achieve this ?
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.