DeepPoseKit VS OpenCV

Or a variation called DeepPoseKit - https://github.com/jgraving/DeepPoseKit which hasn't been as updated as recently but is easier to batch / code.

If you want to do the NN solution, I'd record some videos of the papers being held, toss those into DeepPoseKit to create my landmarks, then use a polygon mask to slice the image.

To get the board corners I estimated them with deepposekit. https://github.com/jgraving/DeepPoseKit

This is what I call a disgusting code : https://github.com/jgraving/deepposekit

Referenceshttps https://www.kaggle.com/datasets/riyaelizashaju/skin-disease-image-dataset-balanced?fbclid=IwAR3wbTp8l5yo_5fx6HAX8Vd2-9cca3khAc8EiBGFObaALfdVid29IuB_rYE https://keras.io/api/applications/vgg/ https://www.tensorflow.org/tutorials/images/cnn?hl=th https://opencv.org/

Data analysis involves scrutinizing datasets for class imbalances or protected features and understanding their correlations and representations. A classical tool like pandas would be my obvious choice for most of the analysis, and I would use OpenCV or Scikit-Image for image-related tasks.

You'll need to plot the predictions. There are a few open source tools to do that, supervision is one you can use (https://github.com/roboflow/supervision) and opencv is another common option (https://github.com/opencv/opencv)

You might be able to achieve this with scripting tools like AutoHotkey or Python with libraries for GUI automation and image recognition (e.g., PyAutoGUI https://pyautogui.readthedocs.io/en/latest/, OpenCV https://opencv.org/).

The OpenCV4NodeJs A.I. library provides an interface for calling OpenCV routines in NodeJS.