1adrianb / binary-human-pose-estimation Goto Github PK

Hi, This demo does not seem to detect multiple people posture ?

Hi, I am trying to install your amazing software on my mac without nvidia gpu, bnn does not build since it tries to find cuda:

CMake Error at /usr/local/Cellar/cmake/3.10.2/share/cmake/Modules/FindCUDA.cmake:682 (message):
Specify CUDA_TOOLKIT_ROOT_DIR
Call Stack (most recent call first):
CMakeLists.txt:10 (FIND_PACKAGE)

How to get the prediction for a new image?
I renamed dataset/mpii/images/07218876.jpg dataset/mpii/images/07218876(copy).jpg
and created a new dataset/mpii/images/07218876.jpg file.
Then I got the prediction for the old dataset/mpii/images/07218871.jpg.
Thank you in advance.

when exec th download-content.lua, the image can not be downloaded, and it can not be opened by web browser too

I got the message below.
Environment: Ubuntu14.04, CUDA7.0, cudnn4.0
Could anyone help me?
Thank you in advance.

$ th main.lua
~/torch/install/bin/luajit: ~/torch/install/share/lua/5.1/nn/Container.lua:67:
In 11 module of nn.Sequential:
~/torch/install/share/lua/5.1/cudnn/BatchNormalization.lua:69: bad argument #1 to 'resizeAs' (torch.CudaTensor expected, got userdata)
stack traceback:
[C]: in function 'resizeAs'
...torch/install/share/lua/5.1/cudnn/BatchNormalization.lua:69: in function <...torch/install/share/lua/5.1/cudnn/BatchNormalization.lua:63>
[C]: in function 'xpcall'
~/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
~/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
~/torch/install/share/lua/5.1/optnet/init.lua:387: in function 'optimizeMemory'
main.lua:27: in main chunk
[C]: in function 'dofile'
~/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00406670
WARNING: If you see a stack trace below, it doesn't point to the place where this error occurred. Please use only the one above.
stack traceback:
[C]: in function 'error'
~/torch/install/share/lua/5.1/nn/Container.lua:67: in function 'rethrowErrors'
~/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
~/torch/install/share/lua/5.1/optnet/init.lua:387: in function 'optimizeMemory'
main.lua:27: in main chunk
[C]: in function 'dofile'
~/torch/install/lib/luarocks

hi,
I have a question about the final block.
In your paper, you mentioned multi-scale should be used in Fig.4e which is your final block. However, Fig.4e only contains 3x3 conv. It doesn't show any pooling or conv with stride operation. I can't understand how to extract multi-scale feature by using it. Could you explain more details about it? Thanks a lot.

liuhao@umaster:~/binary-human-pose-estimation$ th download-content.lua
/home/liuhao/torch/install/bin/luajit: /home/liuhao/torch/install/share/lua/5.1/trepl/init.lua:389: module 'cURL' not found:No LuaRocks module found for cURL
no field package.preload['cURL']
no file '/home/liuhao/.luarocks/share/lua/5.1/cURL.lua'
no file '/home/liuhao/.luarocks/share/lua/5.1/cURL/init.lua'
no file '/home/liuhao/torch/install/share/lua/5.1/cURL.lua'
no file '/home/liuhao/torch/install/share/lua/5.1/cURL/init.lua'
no file './cURL.lua'
no file '/home/liuhao/torch/install/share/luajit-2.1.0-beta1/cURL.lua'
no file '/usr/local/share/lua/5.1/cURL.lua'
no file '/usr/local/share/lua/5.1/cURL/init.lua'
no file '/home/liuhao/.luarocks/lib/lua/5.1/cURL.so'
no file '/home/liuhao/torch/install/lib/lua/5.1/cURL.so'
no file '/home/liuhao/torch/install/lib/cURL.so'
no file './cURL.so'
no file '/usr/local/lib/lua/5.1/cURL.so'
no file '/usr/local/lib/lua/5.1/loadall.so'
stack traceback:
[C]: in function 'error'
/home/liuhao/torch/install/share/lua/5.1/trepl/init.lua:389: in function 'require'
download-content.lua:1: in main chunk
[C]: in function 'dofile'
...uhao/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00405d50

seems the problem is cURL, but I have installed it, have you encounter this? I am the new guy to torch

As far as I am understanding the paper and the implementation, the input image is not binarized, the weight is binarized by bnn binary operation, while the outputs are still real number, am I correct?

I am new to lua and have been trying to use the demo sample. I have follow the instruction and the following show after I run "th main.lua"
home/linux/torch/install/bin/luajit: cannot open <models/human_pose_binary.t7> in mode r at /home/linux/torch/pkg/torch/lib/TH/THDiskFile.c:673
stack traceback:
[C]: at 0x7f4ed56c5450
[C]: in function 'DiskFile'
/home/noble/torch/install/share/lua/5.1/torch/File.lua:405: in function 'load'
main.lua:19: in main chunk
[C]: in function 'dofile'
...oble/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:150: in main chunk
[C]: at 0x00405d50
Thanks.

Hi! Recently I'm trying to do some resherch about binary CNNs models. Luckily I found your paper of "Binarized Convolutional Landmark Localizers for Human Pose Estimation and Face Alignment with Limited Resources" in arXiv. After reading it, I’ve grown a strong interest in the original binarized HG network and the further investigative networks. However, I've noticed that only the “Hierarchical, Parallel & MS Block" HG network model was released in the website. I'm wodering if you can send me those original trained models so i can deploy and compare the inference speed between them.

Thanks a lot.

Is there a chance to release a training script in the near future?

Hi, adrianb,
When I run your evaluation code (mode==eval), the getFileList() function in utils.lua extracts 2958 data. But download-content.lua only provides ten images running demo code (mode==demo). Can you provide all 2958 image data to me? thank you very much.

Hey @1adrianb , could you provide an example of how to train such a binarized convolutional model?

Hi, are there performance benchmarks available for different operating systems? Duration, memory consumption, etc.?

Would be greatly appreciated, thanks!

Hey @1adrianb , could you give me a guide how can i train the model with my own dataset? Is there an example to train it ?