ComfyUI Circle Detection Node 
Important
🚧 This documentation is still under construction. Please note also that the node is not catching all errors e.g. on wrong settings. This is on my to-do list.
Introduction
Next to AI mathematical methods can be used for the detection of objects like a circle. This node is using Hough's Circle Transform for the detection of circles in an given image. The advantage of mathematical methods is that the recognised circles can be virtually predicted. I utilise this to create circle masks.
Goal of the Circle Detection Node
To explain the goal is best done using an example. Take the office under the sea. If I want to move this office in outer space, I have to exchange the content seen in the portholes. By use of the node this can easily done.
Figure 1: Original image
Applying the node and inpainting one gets.
Figure 2: Inpainted image
Usage
After installation the node can be find in the node menu
- Add Node > 🧬 Circle Detection Nodes > 🔬 Circle Detection
For showing data one can use following node
-
Add Node > 🧬 Circle Detection Nodes > 📄 Show Data For the input of data one can use following node
-
Add Node > 🧬 Circle Detection Nodes > ✏️ Input Data
The first node is the node for the circle detection. The second node is for viewing data from the circle detection node.
What the Node Does
Node Input / Output
The input of the node is an image. The output of the node is
- NODE INPUT
- image
- NODE OUTPUT
- image_output
- image_mask
- mask
- data
- help
For showing the data one needs a text/string output node. The help output can be shown with an text/string ouptput node.
Note
help is a feature that should make it in the future easier to work with the node without visiting this documentation.
Settings
List of Settings
- threshold_canny_edge
- threshold_circle_center
- minR
- maxR
- dp
- minDist
- color_tuple_cicles
- color_tuple_bg
- color_tuple_fg
- thickness
- show_circle_center
- numbering
- number_size
Preview of Settings
Figure 2: Circle detection node
Setting color_tuple_cicles
color_tiple_circle sets the color of the circle which marks the setected circle.
Setting exclude_circles
As I wrote earlier mathematical methods produce on each run the same results. One can take advantage of this. After first run one can exclude circles which were detected in the first run.
Version Previews
version 0.0.0.4
Figure 3: Preview of workflow using the circle detection node
version 0.0.0.3
Figure 4: Preview of workflow using the circle detection node
version 0.0.0.2
Figure 5: Preview of workflow using the circle detection node
version 0.0.0.1
Figure 6: Preview of workflow using the circle detection node
Installation
Use the ComfyUI Manager for the installation of the node.
You can also move int the directory ComfyUI/custom_nodes
git clone https://github.com/zentrocdot/ComfyUI_Circle_Detection
Example
Figure 7: Original image
Figure 8: Image with circles detected
To-Do
Troubleshooting in the node programming and sanitizing up the code.
Remarks
The first image created is the one, where on can see, which circles are found. The seond image looks like a mask, but it is not for the moment. It is still an blank image, where the found circles are filled drawn. This changed in version 0.0.0.3.
Open Issues
It is unclear, if the JavaScript event trigger for updating the Show Data window will work.
References
[1] https://docs.opencv.org/3.4/d4/d70/tutorial_hough_circle.html
[2] https://github.com/exectails/comfyui-et_infoutils