I have collected my gesture dataset and I want to train my own model.
I have some probem about the code "self.map = {'draw': 1, 'erase': 0}" in train.ipynb.
Now I have collected seven gestures and they have been put together in one csv file。if I want to train my own gesture model , how do I change this line of code"self.map = {'draw': 1, 'erase': 0}",and where will the trained model be stored? @gursi26

Creating a simple GUI to allow for easier user interaction. Could include functions such as the following:

• Allow users to change pen colour.
• Allow users to change the thickness of the pen.
• Allow users to save their drawings as an image file.

Stylize the README.md file and give it a professional look

• Format heading
• Add badges

The variable dataset is always assigned None. This is because append always returns None since it modifies the list in-place.

Black is an automatic code formatter. While not necessarily fit for everyone's taste, it's PEP8 compliant, can be configured, and has a few additional defaults which can result in a cleaner code. This ensures a uniform coding convention across commits.

https://pypi.org/project/black/

Edit: it also automates the formatting process, so there's less overhead when writing code. It reduces the need to make the code aesthetic and PEP8 compliant.

I think it would make it easier to contribute if there was a standardized code style.
For example, on line 93 (of main.py) there is whitespace before the colon while on line 115 there isn't:

(The latter would be considered the correct one according to PEP8)

Implement the application in flask rather than a simple cv2 window.
Reference article : https://towardsdatascience.com/video-streaming-in-web-browsers-with-opencv-flask-93a38846fe00

The draw function currently detects whether the user's hand is in the "draw" position (only index finger sticking out) and stores the coordinates of the tip of the user's index finger. These coordinates are stored in a list.

In every frame, a blue circle is drawn at each coordinate in the list of coordinates. Due to technical constraints, the program runs at approximately 15-20 FPS and therefore, when the user moves their hand in one quick motion, the line ends up being disjointed, as seen in the image below.

This can be fixed with a function that checks if points drawn in two consecutive frames are far apart, and then appends the coordinate list with a set of intermediate points in between the far apart points.

There is some code duplication in the project. Such an example is:

Improving accuracy of model by using euclidean distance between hand landmarks rather than their raw coordinate value. Gestures will be better recognised on any part of the screen.

The if-statement holds a condition which is always false. It is unclear what it does there. Since it's an operand for an and operator, this means the condition is always false as a whole.

good job!

I want to create some new action with hand tracking,
it seems use data_collection.py to collect gestures datasets,then train the model.

but I do not know how use the code to collect datasets , could you give me some advice?
thanks a lot!

The README says what libraries are required but I think it would be more user friendly if I could just do pip install -r requirements.txt instead of pip installing PyTorch, Mediapipe, and OpenCV one at a time.