This repo is created for the sake of drawing sci-style figures with free softwares (e.g. matplotlib and seaborn).

Currently there are plenty of choices to draw sci-style figures, such as Gnuplot, Origin-Lab. However, based on python, matplotlib are powerful tools to work with powerful abilities and programmable behaviors. With carefully designed style and layout settings, we can plot figures similar to those generated by Origin-Lab, which is a commercial software.

1. Draw Bars

   def draw_bar(
       axes: Axes,
       bar_cfg: Dict[str, Any],
       draw_cfg: Dict[str, Any]
   ) -> Tuple[Figure, Axes]:
       ...

   This function draw a vertical bar plot with the given axes.

   Configurations:

   1. bar_cfg

      • data is a DataFrame contains a set of bar data, including position and error (optional).

      Example:

      	method	AP	err_n	err_p
      0	Redundancy	47.8321	0.4	0.3
      1	Isometric	47.2445	0.2	0.4
      2	Random	45.53119	0.2	0.3
      3	SAG	44.64032	0.2	0.2

      • error is an optional dict contains error type and corresponding to the keys in bar_cfg. Default None.

      error: # dict
          err_pos: err_p  # positive error
          err_neg: err_n  # negative error, optional if errors are symmetric

      • x_axis is the column name of the x axis in data, e.g. method.
      • y_axis is the column name of the y axis in data, e.g. AP.

   2. draw_cfg

      • bar contains style settings of the bar in the figure, referenced to matplotlib.axes.Axes.bar, type: dict

      bar: # dict
          width: 0.45         # width of each rectangular
          align: center       # alignment to x-ticks
          edgecolor: black
          linewidth: 0.8

      • bar_label contains style settings of the label of each rectangular in the bar plot, referenced to matplotlib.axes.Axes.bar_label, type dict

      bar_label: # dict
          fmt: "%.1f"
          label_type: edge
          fontsize: small

      • error_bar contains style settings of the error bar, referenced to matplotlib.axes.Axes.errorbar, type dict

      error_bar: # dict
          ecolor: black
          elinewidth: 0.8
          markeredgewidth: 0.8

      • colormap reference to Sec. ColorMap
      • color is a list of colors, this will override the colormap setting, type: List[str]

2. Draw Heatmap

3. Draw Lines

1. Basic Bins

   Path: examples/base_bins/vis_base_bins.py

   

2. Heatmap

   Path: examples/lr_curves/vis_xxx.py

3. Line with Markers

4. Loss Learning Curve

5. LR Learning Curve

1. Seaborn
2. Matplotlib
3. https://matplotlib.org/cheatsheets/cheatsheets.pdf