Tensorflow recently released their new object detection api Is there any way to run this on windows? The directions apear to be for linux.
Yes, you can run the Tensorflow Object Detection API on Windows. Unfortunately it is a bit tricky and the official documentation does not reflect that appropriately. I used the following procedure:
Install Tensorflow natively on Windows with Anaconda + CUDA + cuDNN. Note that TF 1.5 is now built against CUDA 9.0, so make sure you download the appropriate versions.
Then you clone the repository and build the Protobuf files as described in the tutorial, but beware, there is a bug in Windows Protobuf 3.5, so make sure you use version 3.4.
cd [TF-models]\research
protoc.exe object_detection/protos/*.proto --python_out=.
Finally, you need to build and install the packages with
cd [TF-models]\research\slim
python setup.py install
cd [TF-models]\research
python setup.py install
If you get the exception error: could not create 'BUILD': Cannot create a file when that file already exists here, delete the BUILD file inside first, it will be re-created automatically
And make the built binaries available to your path python path, or simply copy the directories slim and object_detection to your [Anaconda3]/Lib/site-packages directory
To see everything put together, check out our Music Object Detector, which was trained on Windows and Linux.
We don't officially support the Tensorflow Object Detection API, but some external users have gotten it to work.
Our dependencies are pillow, lxml, jupyter, matplotlib and protobuf compiler. You can download a version of the protobuf compiler here. The remaining dependencies can be installed with pip.
As I said on the other post, you can use your local GPU in windows, as Tensorflow supports GPU on python.
And here is an example.
Unfortunately, Tensoflow does not support tensorflow-serving on windows. Also as you said Nvidia-Docker is not supported on windows. Bash on windows has no support for GPU either. So I think this is the only easy way to go for now.
The below tutorial was build specifically for using the Tensorflow Object Detection API on Windows. I've successfully used the below tutorial many times:
https://github.com/EdjeElectronics/TensorFlow-Object-Detection-API-Tutorial-Train-Multiple-Objects-Windows-10
Related
So I have been writing a Python program that utilizes OpenCV for Windows. It's mostly just a project to learn how to use Vision-based Machine Learning. I've gotten the project working with my CPU, and while it "functions" it is abysmally slow, so I wanted to try working with the GPU version instead. Unfortunately, there isn't a python package that has OpenCV with Cuda-enabled GPU functionality (from what I could tell).
So after researching, I found that in order to do what I wanted, I had to compile the OpenCV source code with Cmake. So I set out to do so with the help of this guide which seemed to work (it compiled at least). But now i'm running into the issue of:
I don't actually know how to import this newly built package into my project's venv. I've tried moving the cv2.cp37-win_amd64.pyd into the project's venv site packages, i've tried moving the entire build folder into the project and directly importing it, but neither actually worked... so I'm a little at a loss on what to do to get OpenCV with GPU enabled CUDA working in my project.
EDIT:
I followed the guide that Miki linked, including installing Anaconda and all that jazz. The package imports correctly when I use:
Following #Miki's suggestion, I went through the guide that they linked, following all of the processes, testing that OpenCV was in fact being built correctly with:
set path=%openCvBuild%\install\x64\vc16\bin;%path%
python -c "import cv2; print(f'OpenCV: {cv2.__version__} for python installed and working')"
But it still isn't showing up in the python import in conda, despite being in the env's site-packages folder.
I want to use opencv-python4.4 to do inference for YOLOv3. But it uses CPU by default. I have read this instruction here, saying the new version opencv4.2.0 has already supported GPU. But the instruction is only for Ubuntu but mine is Windows10.
I tried to Could you please walk me through how to pull this off in Wondows10, python? As I directly used it, the error occurred:
cv::dnn::dnn4_v20191202::Net::Impl::setUpNet DNN module was not built with CUDA backend; switching to CPU
It seems that I have to manually compile opencv but isn't it for Cpp? But I only want to implement it on python. Could anyone give me some instructions?
You will indeed need to build OpenCV yourself.
While you're using the Python bindings for OpenCV, the OpenCV library itself is written in C++ instead of Python. That also explains how OpenCV can use CUDA, another C++ library to access NVidia GPU's.
The instructions you linked are from a person not associated with OpenCV, who admits to an anti-Windows bias. That means those instructions are not useful to you.
These instructions do cover the Windows build.
Summary:
Use Visual Studio 2019 (Latest C++ compiler from Microsoft) and CMake
Download & install OpenCV
Download & install CUDA and cuDNN
Download & install Anaconda3 and use it as default Python
Set environment variables so CMake can find your installed libraries
Set the environment variables and options so CMake knows that you also want the Python bindings for OpenCV
Use CMake to create .sln file for Visual Studio
Open .sln in Visual Studio and build it.
This generates the cv2.cp37-win_amd64.pyd file you need.
I have figured it out... both
-DBUILD_opencv_python3=ON
-DBUILD_opencv_python2=OFF
must be specified, otherwise the to be built will not include any python module at all.
I have never used tensorflow (or python). I installed Python 3.5.2 from https://www.python.org/downloads/. Then I followed instructions to install tensorflow according to this page, for a Windows 64-bit installation:
https://www.tensorflow.org/versions/r0.12/get_started/os_setup.html#download-and-setup
I followed the steps for Pip installation on windows as a CPU-only install. Then I followed Test the TensorFlow installation/ Run TensorFlow from command line which also seemed mostly fine.
But I hit a brick wall when I try the section Run a TensorFlow demo model. Specifically, the command python -m tensorflow.models.image.mnist.convolutional returns a Error while finding spec for 'tensorflow.models.image.mnist.convolutional' (ImportError: No module named 'tensorflow.models').
I don't really understand the installation nor what I'm missing. Can you tell me what I've screwed up?
Tensorflow models have been moved to a different github repository.
Therefore, they wont be available out of the box anymore when TF is installed.
You can find the file right here:
https://github.com/tensorflow/models/blob/master/tutorials/image/mnist/convolutional.py
You can download this file and execute it manually.
Sorry for out of date documentation.
To run the example the way you've described I think you'd have to add the tensorflow directory (or directory containing all 3rd party packages) to the PYTHONPATH environment variable as in the answer to this question or be in a directory where python can find the tensorflow module. Alternatively, you can run the example using:
python path/to/convolutional.py
from the command line which passes the script to python.exe and executes it. Make sure to either put in the full path or cd to the directory containing the script.
Using Anaconda as mentioned in your second link might be the best bet as setting PYTHONPATH will be taken care of for you for 3rd party packages installed either using conda or pip.
I tried to install a hoomd_script molecular dynamics software on my imac (it's imac pro before 2009, the system is OS X El captain v10.11.3). I have successfully compiled this to iMac, but when I import this hoomd_script in Python 2.7.12, Python crashes completely and I get the error:
Illegal instruction: 4.
I have installed all the prerequisites packages (including boost, sphinx, git, mpich2, numpy, cmake, pkg-config, sqlite) using conda.
I applied python -vc 'hoomd_script' to test, and the result is here. I tried to reinstall all the packages including conda and recompile the hoomd, but nothing changed. I wonder how can I fix this. Thanks!
As stated on the HOOMD-blue web page, the conda builds require a CPU capable of AVX instructions (2011 or newer). The illegal instruction results because you are trying to execute an instruction that your processor does not support.
Compiling hoomd from a clean build directory on your system should result in a binary that your system can execute. Note that conda provided prerequisite libraries are difficult to work with: I recommend using macports or homebrew.
I am new to Python and I want to install VLFeat on Ubuntu (13.04).
I am using Eclipse 3.8. For Python, I have installed the PyDev extension on Eclipse.
I have installed Numpy, but I don't know how to install VLFeat. I tried to use their website, but I can't get anything for Python. I have downloaded packages, but I don't know how to install them.
The Menpo Project provides a wrapper around VLFeat: it's called cyvlfeat.
To install cyvlfeat, we strongly suggest you use conda:
conda install -c menlo cyvlfeat
If you don't want to use conda, your
mileage will vary. In particular, you must satisfy the
linking/compilation requirements for the package, which include the
vlfeat dynamic library.
In other words, the nice thing about installing with conda is that it will install (and link) VLFeat dependencies as well.
It may not include all functionalities of VLFeat. Current State as of March 2017:
fisher
fisher
generic
set_simd_enabled, get_simd_enabled, cpu_has_avx, cpu_has_sse3, cpu_has_sse2
get_num_cpus,
get_max_threads, set_num_threads, get_thread_limit
hog
hog
kmeans
kmeans
kmeans_quantize
ikmeans, ikmeans_push
hikmeans, hikmeans_push
sift
dsift
sift
Relevant reading
Dev blog by Simmi Mourya. This includes descriptions and usage examples.
More alternatives:
vlfeat-ctypes: minimal VLFeat interface for python
pyvlfeat fork by jchazalon: A high-level Python wrapper around a subset of the VLFeat library [more recently updated than the original]
Note about the IDE
Installing python packages should be independent of the IDE (Eclipse + PyDev, in the OP case), as long as the interpreter and libraries paths are correctly set up.
Note about conda
It is not required to install the Anaconda distribution in order to use conda. The much lighter Miniconda is enough.
Assuming you are getting VLFeat from the Python Package Index, the instructions are
Download distribution, extract it, get to command prompt and type:
$ python setup.py install