ComfyUI Circle Detection Node

[!IMPORTANT]

<p align="justify">🚧 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.</p>

Introduction

<p align="justify">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.</p>

Goal of the Circle Detection Node

<p align="justify">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.</p>

Figure 1: Original image

Using the capabilities of the node together with a painted mask and applying inpainting one gets.

Figure 2: Inpainted image

Usage

<p align="justify">After installation the node can be find in the node menu by searching for the entry <code>🧬 Circle Detection Nodes</code>.</p>

• 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

<p align="justify">The first node is the node for the circle detection. The second node is for viewing data from the circle detection node. The third node is for the input of the optional data.</p>

What the Node Does

Basic Feature

<p align="justify">The basic feature of the node is the detection of circles in images. The second feature is the creation of a mask from the detected circles.</p>

Node Input / Output

The input of the node is an image. The output of the node is

• NODE INPUT
  • image (required)
  • color_tuple_bg (optional)
  • color_tuple_fg (optional)
  • exclude_circles (optional)
• NODE OUTPUT
  • image_output
  • image_mask
  • standard_mask
  • inverted_mask
  • show_terminal_data
  • circle_detection_help

The image output is an image where the detected circles are marked. The image mask is a colored image consisting of background and foreground. A standard mask as well as an inverted mask can be used for a further processing of the image.

For showing the data one needs a text/string output node. The help output can be shown with an text/string ouptput node.

[!NOTE]

<p align="justify">help is a feature that should make it in the future easier to work with the node without visiting this documentation.</p>

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

<p align="justify">The preview images of the first four versions in form of a workflow are shown below.</p>

Figure 2: Circle detection node

Setting color_tuple_cicles

<p align="justify">color_tiple_circle sets the color of the circle which marks the setected circle.</p>

Setting exclude_circles

<p align="justify">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.</p>

Setting numbering

The numbering of the detected circles can be enabled/disabled by this setting.

Special Feature

<p align="justify">After first run detected circles can be excluded by use of the optional input connector.</p>

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

<code>git clone https://github.com/zentrocdot/ComfyUI_Circle_Detection</code>

Example

<img src="./images/ComfyUI_0001.jpeg" alt="button panel" width="512"> <p><i>Figure 7: Original image</i></p> <img src="./images/ComfyUI_0002.jpeg" alt="button panel" width="512"> <p><i>Figure 8: Image with circles detected</i></p>

To-Do

Troubleshooting in the node programming and sanitizing up the code.

Remarks

<p align="justify">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.</p>

Open Issues

<p align="justify">It is unclear, if the JavaScript event trigger for updating the Show Data window will all the time work. This has to be checked.</p>

If a circle is not perfect or a circle is deformed or malformed such a circle cannot be detected with his contour. I have to develop a node, which can identify such malformed/deformed circles.

The last one is a parallel activity for the detection of ellipses.

References

[1] https://docs.opencv.org/3.4/d4/d70/tutorial_hough_circle.html

[2] https://github.com/exectails/comfyui-et_infoutils

[3] https://github.com/rgthree/rgthree-comfy

[4] https://github.com/pythongosssss/ComfyUI-Custom-Scripts