Helper Module for Deep Learning.

Common functions to display a dynamic board.

class pynet.plotting.board.Board(port=8097, host='http://localhost', env='main', display_pred=False, prepare_pred=None)[source]¶

Define a dynamic board.

It can be used to gather interesting plottings during a training.

__init__(port=8097, host='http://localhost', env='main', display_pred=False, prepare_pred=None)[source]¶

Initilaize the class.

Parameters

port: int, default 8097

the port on which the visdom server is launched.

host: str, default ‘http://localhost’

the host on which visdom is launched.

env: str, default ‘main’

the environment to be used.

display_pred: bool, default False

if set render the predicted images.

prepare_pred: callable, defaultt None

a function that transforms the predictions into a Nx1xXxY or Nx3xXxY array, with N the number of images.

update_plots(data, epoch)[source]¶

Update/create plots from the input data.

Parameters

data: dict

the name and new value of the plot to be updated.

epoch: int

the current epoch.

pynet.plotting.board.update_board(signal)[source]¶

Callback to update visdom board visualizer.

Parameters

signal: SignalObject

an object with the trained model ‘object’, the emitted signal ‘signal’, the epoch number ‘epoch’ and the fold index ‘fold’.