Dashboard
Simple dash board to visualize research progress.
Right now the dashboard is table based. Each row in a table is an entry in the dashboard. The entry is parsed automatically, the fist integer / float value is plotted. Any string preceding this entry is is used as a label. Any preceding date is used as timestamp.
Special names are
- 's:\w*=.*' specify table settings (see TableSettings)
s:\w*enables a boolean flag
- 'i:' ignore this entry
- 'b:' baseline visualized as a horizontal line
Every sheet of a table will be visualized in a tab. The tabs can be sorted using the s:priority=?? tag where 10 is the default priority.
You can customize what is shown in the table by settings the number of top results s:keep_top=??, and the number of most recent results s:keep_last=??.
By default the top entries are the ones with the highest value, if you'd like to visualize the lowest use s:lower_better.
The y-axis label is specified as s:axis=.... To hide a sheet use s:hide.
Here is a simple example of a table
Sheet 1 - COCO
| s:axis=mAP | ||
|---|---|---|
| b:some baseline | 1/2/2020 12:00:00 | 10.1 |
| b:other baseline | 1/2/2020 12:00:00 | 20.5 |
| first model | 1/3/2020 | 22.5 |
| second model | 1/4/2020 | 23.5 |
| tenth model | 1/10/2020 | 30.0 |
Sheet 2 - KITTI
| s:axis=error | s:lower_better | s:keep_top=10 |
|---|---|---|
| b:some baseline | 1/2/2020 12:00:00 | 20.5 |
| b:other baseline | 1/2/2020 12:00:00 | 15.5 |
| first model | 1/3/2020 | 17.7 |
| second model | 1/4/2020 | 13.2 |
| tenth model | 1/10/2020 | 0.001 |