Parallelism
- Vojtěch Kováčik
Parallelism enables parallel processing in several branches and their results are then combined.
For instance, you can use 2 different modules, each in a separate branch of the Parallelism, and their results will be combined at the end of the branching in the following way:
Evaluation result OK/NG - If one branch provides a result ok and the other ng , the result will be ng , which is a combination of both branches (boolean operator AND).
Detected rectangles and heatmaps - Rectangles and heatmaps from all branches will appear together in the joined output.
Image - The results from all branches are mapped to the original image (image before parallelism). This works also with using preprocessing (cut, scale) in the branches when results will be adjusted to fit the original image.
In order to use Parallelism, firstly, add it from the Modules tab, then you have 3 options to move modules around its branches:
Use the + button to add a module directly to the desired branch.
Use the arrow to move a module to Parallelism branches.
Once a module is added to a branch, you can use a side arrow to move a module to the right or left.
In the final position we have two Modules. An evaluation from Classifier and an evaluation from Detector. In the resulting image we are going to see rectangles from both modules with their corresponding classes.