MorVision

Note that this is only tested on Ubuntu and probably does NOT run on windows

!!!Do NOT update models repository here without testing it first!!!

Installation

Tensorflow

Look at guides here.

Sources are included in the tf folder and can be installed by following this guide.

MKL Support

Install from sources and add --config=mkl to the bazel build

Please see tensorflow.org for more information MKL can be installed by adding deb https://apt.repos.intel.com/mkl all main to /etc/apt/sources.list/d/intelproducts.list (Note that /intelpython, /ipp, /tbb, and /daal also exist on apt.repos.intel.com, although /intelpython is down as of April 16th, 2018, see intel.com)

Object Detector

1. run

sudo apt-get install protobuf-compiler python-pil python-lxml python-tk
sudo pip install matplotlib jupyter

2. run

protoc models/research/object_detection/protos/*.proto --python_out=./models/research

3. run or add absolute paths to .bashrc

cd models/research;export PYTHONPATH=$PYTHONPATH:`pwd`:`pwd`/slim;cd ../..

4. test by running

python models/object_detection/builders/model_builder_test.py

Annotation

Just run npm start inside of the LabelMeAnnotationTool folder

This starts a webserver on localhost:8080

Creating Data

python make_model/make_model.py

Training

Running Tensorflow

python models/research/object_detection/train.py --train_dir=./temp --pipeline_config_path=`pwd /make_model/embedded_ssd_mobilenet_v1_coco.config`

Running TensorBoard

tensorboard --logdir=./temp

Note that to see this open browser to localhost:6006

Running Evaluation

python models/research/object_detection/eval.py --eval_dir=./tempEval --pipeline_config_path=/home/elias/Desktop/web/morvision/make_model/embedded_ssd_mobilenet_v1_coco.config