Uses a Web UI to visualize plots and statistics with the data generated during feaure-extractor training or evaluation.
Visualize via Web, data obtained from an input plugin, the data obtained via the input plugin may contain batch or real-time results of multiple feature-extractor trainers or evaluators (from now on called processes). By default uses a Sqlite input plugin.
It uses multiple output visualization plugins, each of which generate a new element in the feature-extractor dashboard views of the feature-extractor processes. By default uses output plugins for: real-time MSE plot during training and batch calculated MSE plot from the evaluation of a pre-trained feature-extractor on a validation dataset.
The visualizer uses a JSON configuration file for setting the Web service parameters and the configuration of the input and output plugins.
To install the package via PIP, use the following command:
pip install -i https://test.pypi.org/simple/ visualizer
Also, the installation can be made by clonning the github repo and manually installing it as in the following instructions.
On Linux use the .sh scripts, on windows use the .bat scripts.
- Clone the GithHub repo:
git clone https://github.com/harveybc/visualizer
- Change to the repo folder:
cd visualizer
- Install requirements.
pip install -r requirements.txt
- Install python package (also installs the console command data-trimmer)
python setup.py install
- Add the repo folder to the environment variable PYTHONPATH
- Create a test database (use migrate.bat on windows)
scripts/migrate.sh
- Populate the test database (use test_data_seed.bat on Windows)
scripts/test_data_seed.sh
- (Optional) Perform tests
python setup.py test
- (Optional) Generate Sphinx Documentation
python setup.py docs
For ease of use, a script for setting the environment variables and executing the app is included, it must be executed in the root feature-extractor directory, where the scripts are located:
- For Linux and Mac:
visualizer.sh
- For Windows:
visualizer.bat
The Web intreface can be accessed by default at:
localhost:5000
The default port can be modified by setting the FLASK_RUN_PORT environment variable or bly using the --port argument to the flask run command.
A default user us created with the username: "test", and password: "pass", please delete this user once you have created another one.
The visualizer uses a configuration file located in the visualizer/visualizer directory that sets the Web service parameters and the configuration of the input and output plugins.
The following is the default JSON configuration file:
{
"input_plugin": "vis_input_sqlite",
"input_plugin_config": {
"filename": "test/db/plots.sqlite",
"tables": [
{
"table_name": "training_progress",
"fields": [
"mse",
"mae",
"r2"
]
},
{
"table_name": "validation_stats",
"fields": [
"mse",
"mae",
"r2"
]
},
{
"table_name": "validation_plots",
"fields": [
"original",
"predicted"
]
}
]
},
"output_plugin": "vis_output",
"output_plugin_config": {
"dashboard": [
{
"table_name": "training_progress",
"online": true,
"delay": 3,
"points": 800,
"title": "Progress of Last Training Process",
"fields": [
"mse"
]
},
{
"table_name": "validation_plots",
"title": "Validation Data Plot"
},
{
"table_name": "validation_stats",
"title": "List of Feature Extractor Stats on Validation Data"
}
],
"views": [
{
"table_name": "training_progress",
"title": "Progress of Training Process",
"online": true,
"delay": 3,
"points": 1600,
"fields": [
"mse",
"mae",
"r2"
]
},
{
"table_name": "validation_plots",
"title": "Validation Data Plot",
"online": false,
"fields": [
"original",
"predicted"
]
},
{
"table_name": "validation_stats",
"title": "Feature Extractor Stats on Validation Data",
"online": false,
"fields": [
"mse",
"mae",
"r2"
]
}
]
}
}
.
.