AndreyFilippov committed changes to the Elphel git project :
Implemented MPO export for EO/LWIR dual-quad rig

AndreyFilippov committed changes to the Elphel git project :
Implementing batch mode for EO/LWIR images

AndreyFilippov committed changes to the Elphel git project :
AUX (LWIR) extrinsic adjustments with main camera ground truth

AndreyFilippov committed changes to the Elphel git project :
compositeScan() with multi-plane (FG/BG) results

AndreyFilippov committed changes to the Elphel git project :
Ground truth from EO to LWIR

Oleg Dzhimiev committed changes to the Elphel git project :
switched to flir's iron palette

AndreyFilippov committed changes to the Elphel git project :
minor cleanup

AndreyFilippov committed changes to the Elphel git project :
refactoring, alignment with LWIR sensors

AndreyFilippov committed changes to the Elphel git project :
Autoranging for lwir images/texture tiles

AndreyFilippov committed changes to the Elphel git project :
splitting LPF sigma for Bayer/mono

AndreyFilippov committed changes to the Elphel git project :
Getting processed LWIR images

AndreyFilippov committed changes to the Elphel git project :
monochrome mode of the tile processor - debugging

AndreyFilippov committed changes to the Elphel git project :
converting texture processing to work with monochrome too

AndreyFilippov committed changes to the Elphel git project :
next snapshot working on monochrome mode in TP

‎Setup (guide)

← Older revision Revision as of 23:32, 2 July 2019 (17 intermediate revisions by the same user not shown)Line 1: Line 1:  ==Requirements== ==Requirements== −* Kubuntu 16.04.4 LTS+* Kubuntu 16.04 LTS  ==Setup (guide)== ==Setup (guide)==  Just follow: Just follow: −* [http://www.python36.com/install-tensorflow141-gpu/ '''this guide'''] (Ubuntu 16.04 64-bit, CUDA 9.1, cuDNN 7.1.2, python3) or+* The [[Tensorflow_with_gpu#Walkthrough_for_CUDA_10.1_.2820190602.29|'''walkthrough''']] in the bottom is for CUDA 10.1, cuDNN 7.6.1, python3 −* this [http://www.python36.com/how-to-install-tensorflow-gpu-with-cuda-9-2-for-python-on-ubuntu/ '''newer one'''] (Ubuntu 16.04 64-bit, CUDA 9.2, cuDNN 7.1.4, python3)+* [http://www.python36.com/how-to-install-tensorflow-gpu-with-cuda-9-2-for-python-on-ubuntu/ '''This guide'''] (Ubuntu 16.04 64-bit, CUDA 9.2, cuDNN 7.1.4, python3)  +* [http://www.python36.com/install-tensorflow141-gpu/ '''This guide'''] (Ubuntu 16.04 64-bit, CUDA 9.1, cuDNN 7.1.2, python3)     ==Setup (some details)== ==Setup (some details)== Line 130: Line 131:    # Solution:   # Solution:    <b>~$ sudo pip3 install setuptools --upgrade</b></font>   <b>~$ sudo pip3 install setuptools --upgrade</b></font>  +  +==Walkthrough for CUDA 10.1 (20190602)==  +  +===Install CUDA===  +* In this [https://www.tensorflow.org/install/gpu guide] there's a [https://developer.nvidia.com/cuda-toolkit-archive link to CUDA toolkit].  +** That toolkit (CUDA Toolkit 10.1 update1 (May 2019)) also updated the system driver to 418.67  +** Reboot  +===Install cuDNN===  +* Have to have an account with NVIDIA - downloaded [https://developer.nvidia.com/rdp/cudnn-download#a-collapse761-101 cuDNN v7.6.1 (June 24, 2019), for CUDA 10.1]  +  +===Option 1: installing tensorflow from source===  +Basically, [https://www.tensorflow.org/install/source '''this guide'''], some key notes:  +* [https://www.tensorflow.org/install/source#install_bazel Install bazel] - version 0.25.2 (newer will not work)  +* To build, read [https://www.tensorflow.org/install/source#download_the_tensorflow_source_code this link]:  + git clone https://github.com/tensorflow/tensorflow.git  + cd tensorflow  + git checkout r1.14  + ./configure  +   + bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package  + # 4-5 hours later  + ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg  + sudo pip3 install /tmp/tensorflow_pkg/tensorflow-[Tab]  +  +* Testing:  + ~$ python3  + >>> import tensorflow as tf  + >>> hello = tf.constant('Hello, World!')                                                                                                                                                                             + >>> sess = tf.Session()  +  +===Option 2: using docker===  +Follow [https://www.tensorflow.org/install/docker '''this guide''']. Key notes:  +* Tensorflow docker image requires nvidia docker image, nvidia docker image requires ''apt install nvidia-docker2'', ''nvidia-docker2'' requires ''apt install docker-ce'':  + - https://github.com/NVIDIA/nvidia-docker  + - https://docs.docker.com/install/linux/docker-ce/ubuntu/  +  +* Test run:  + # Test 1: GPU support inside container:  + sudo docker run --runtime=nvidia --rm nvidia/cuda:10.1-base nvidia-smi  + # Test 2: Test all together  + sudo docker pull tensorflow/tensorflow:latest-gpu-py3-jupyter  + sudo docker run --runtime=nvidia -it --rm tensorflow/tensorflow:latest-gpu-py3-jupyter python -c "import tensorflow as tf; tf.enable_eager_execution(); print(tf.reduce_sum(tf.random_normal([1000, 1000])))"  + # Test 3: Run a local script (and include a local dir) in contatiner:  + https://www.tensorflow.org/install/docker Oleg

‎Option 2: using docker

← Older revision Revision as of 23:13, 2 July 2019 (9 intermediate revisions by the same user not shown)Line 130: Line 130:    # Solution:   # Solution:    <b>~$ sudo pip3 install setuptools --upgrade</b></font>   <b>~$ sudo pip3 install setuptools --upgrade</b></font>  +  +==Walkthrough for CUDA 10.1 (20190602)==  +  +===Install CUDA===  +* In this [https://www.tensorflow.org/install/gpu guide] there's a [https://developer.nvidia.com/cuda-toolkit-archive link to CUDA toolkit].  +** That toolkit (CUDA Toolkit 10.1 update1 (May 2019)) also updated the system driver to 418.67  +** Reboot  +===Install cuDNN===  +* Have to have an account with NVIDIA - downloaded [https://developer.nvidia.com/rdp/cudnn-download#a-collapse761-101 cuDNN v7.6.1 (June 24, 2019), for CUDA 10.1]  +  +===Option 1: installing tensorflow from source===  +Basically, [https://www.tensorflow.org/install/source '''this guide'''], some key notes:  +* [https://www.tensorflow.org/install/source#install_bazel Install bazel] - version 0.25.2 (newer will not work)  +* To build, read [https://www.tensorflow.org/install/source#download_the_tensorflow_source_code this link]:  + git clone https://github.com/tensorflow/tensorflow.git  + cd tensorflow  + git checkout r1.14  + ./configure  +  + bazel build --config=opt --config=cuda //tensorflow/tools/pip_package:build_pip_package  + # 4-5 hours later  + ./bazel-bin/tensorflow/tools/pip_package/build_pip_package /tmp/tensorflow_pkg  + sudo pip3 install /tmp/tensorflow_pkg/tensorflow-[Tab]  +  +===Option 2: using docker===  +Follow [https://www.tensorflow.org/install/docker '''this guide''']. Key notes:  +* Tensorflow docker image requires nvidia docker image, nvidia docker image requires ''apt install nvidia-docker2'', ''nvidia-docker2'' requires ''apt install docker-ce'':  + - https://github.com/NVIDIA/nvidia-docker  + - https://docs.docker.com/install/linux/docker-ce/ubuntu/  +  +* Test run:  + # Test 1: GPU support inside container:  + sudo docker run --runtime=nvidia --rm nvidia/cuda:10.1-base nvidia-smi  + # Test 2: Test all together  + sudo docker pull tensorflow/tensorflow:latest-gpu-py3-jupyter  + sudo docker run --runtime=nvidia -it --rm tensorflow/tensorflow:latest-gpu-py3-jupyter python -c "import tensorflow as tf; tf.enable_eager_execution(); print(tf.reduce_sum(tf.random_normal([1000, 1000])))"  + # Test 3: Run a local script (and include a local dir) in contatiner:  + https://www.tensorflow.org/install/docker Oleg

AndreyFilippov committed changes to the Elphel git project :
more editing for lwir/monochrome, Bayer images are OK

← Older revision Revision as of 22:04, 2 July 2019 (2 intermediate revisions by the same user not shown)Line 130: Line 130:    # Solution:   # Solution:    <b>~$ sudo pip3 install setuptools --upgrade</b></font>   <b>~$ sudo pip3 install setuptools --upgrade</b></font>  +  +==Walkthrough for CUDA 10.1==  +  +===CUDA===  +* In this [https://www.tensorflow.org/install/gpu guide] there's a [https://developer.nvidia.com/cuda-toolkit-archive link to CUDA toolkit].  +** That toolkit (CUDA Toolkit 10.1 update1 (May 2019)) also updated the system driver to 418.67  +** Reboot  +===cuDNN===  +* Have to have an account with NVIDIA - downloaded [https://developer.nvidia.com/rdp/cudnn-download#a-collapse761-101 cuDNN v7.6.1 (June 24, 2019), for CUDA 10.1] Oleg

AndreyFilippov committed changes to the Elphel git project :
Adding monochrome mode to CLT, snapshot tested with Bayer images

Oleg Dzhimiev committed changes to the Elphel git project :
php does not like ../?