1adrianb / human-pose-estimation Goto Github PK

I have found it rather tedious to deploy torch and its dependencies on my computer.
Here is a dockerfile I wrote to do the job pretty much automatically (note that it requires the nvidia wrapper to docker):

https://gist.github.com/pixelou/e06d8b049001254575d745be2b2b0402

FROM nvidia/cuda:8.0-cudnn5-devel

WORKDIR /root

RUN apt-get update
RUN apt-get upgrade -y
RUN apt-get install -y --no-install-recommends openssl ca-certificates \
	git-core cmake curl wget unzip gfortran libreadline-dev ncurses-dev \
	libgoogle-glog-dev libboost-dev libboost-thread-dev libopenblas-dev

# Core Torch packages
RUN git clone https://github.com/torch/luajit-rocks.git
RUN cd luajit-rocks && mkdir build && cd build \
    && cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Release \
             -DWITH_LUAJIT21=ON -DLUAJIT_ENABLE_LUA52COMPAT=ON .. \
    && make && make install
RUN luarocks install torch
RUN luarocks install trepl

# Common torch package
RUN luarocks install luafilesystem
RUN luarocks install --server=http://luarocks.org/dev luash
RUN luarocks install nn
RUN luarocks install optnet
RUN luarocks install xlua
RUN luarocks install cudnn

# Extra torch package
RUN apt-get install -y --no-install-recommends libjpeg-dev libpng-dev \
	graphicsmagick-imagemagick-compat
RUN luarocks install image
RUN apt-get install -y --no-install-recommends python2.7-dev python-numpy
RUN git clone https://github.com/facebook/thpp.git
RUN cd thpp/thpp && mkdir build && cd build \
    && cmake -DCMAKE_INSTALL_PREFIX=/usr -DCMAKE_BUILD_TYPE=Release \
       -DNO_THRIFT=ON -DNO_FOLLY=ON -DNO_TESTS=ON .. \
    && make && make install
RUN luarocks install fbpython

# Pull project
RUN git clone https://github.com/1adrianb/human-pose-estimation

CMD /bin/bash

deploy like so:

curl -O https://gist.githubusercontent.com/pixelou/e06d8b049001254575d745be2b2b0402/raw/9585f22ba540a2fb07b54c4c3b8e314a53010199/Dockerfile
nvidia-docker build -t humposest -f ./Dockerfile
nvidia-docker run -it humposest bash

Hi, thanks for sharing the code of your work.
I have some questions about your method.

Have you tried end-to-end learning (joint learning of detector and regression network)? If so, can you tell me the performance and hyper-parameters (e.g., learning rate)?

Also, how the performance change if deconv layer is initialized by random gaussian distribution( not initialized as bilinear upsampling)?

Hello author
did you use Gaussian filter to draw ellipse shape. How can I achieve this?

thank you

I tried to evaluate the LSP dataset but I got this error:

In 1 module of nn.Sequential:
In 2 module of nn.ConcatTable:
In 1 module of nn.Sequential:
/root/torch/install/share/lua/5.1/nn/THNN.lua:110: bad argument #3 to 'v' (cannot convert 'struct THFloatTensor *' to 'struct THDoubleTensor *')
stack traceback:
        [C]: in function 'v'
        /root/torch/install/share/lua/5.1/nn/THNN.lua:110: in function 'SpatialConvolutionMM_updateOutput'
        /root/torch/install/share/lua/5.1/nn/SpatialConvolution.lua:96: in function </root/torch/install/share/lua/5.1/nn/SpatialConvolution.lua:92>
        [C]: in function 'xpcall'
        /root/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
        /root/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function </root/torch/install/share/lua/5.1/nn/Sequential.lua:41>
        [C]: in function 'xpcall'
        /root/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
        /root/torch/install/share/lua/5.1/nn/ConcatTable.lua:11: in function </root/torch/install/share/lua/5.1/nn/ConcatTable.lua:9>
        [C]: in function 'xpcall'
        /root/torch/install/share/lua/5.1/nn/Container.lua:63: in function 'rethrowErrors'
        /root/torch/install/share/lua/5.1/nn/Sequential.lua:44: in function 'forward'
        main.lua:100: in main chunk
        [C]: in function 'dofile'
        /root/torch/install/lib/luarocks/rocks/trepl/scm-1/bin/th:145: in main chunk
        [C]: at 0x00406670

Do you know what causes it? Thanks

I`m using torch7 to implement my work.
I need to initialize the wieght of deconv layer with bilinear upsampling filter as yours in torch7.
Can you tell me how you did it?

hello authors,

Sorry i want to ask a question related to Multi-person detection.
I want to ask wheter the COCO person dataset is specialized for multi-person?
i have trained a simple model using FCN for face keypoint regression. Can you please tell me how we can get heatmaps of muliple faces in the bottom-up approach. My current model takes input of 96x96x1 image and gives 96x96x15 size heatmaps for 15 keypoints. I trained my model using datset consisting of images with single face. Do i need the datset with multiple faces? and do I need bouding box information or mask information too?

Please give me your advice
thank you