A mouse is a data input device that helps to point and to interact with whatever that's being pointed. In a hardware device like mouse we can still face latency related issues and in rare cases it affects the reliability of the system. Because the mouse will have a durability time within which is functional and after its durability time we've to vary the mouse. Different techniques are being proposed for gaining necessary information/data for the recognition of gestures. Some models work with special devices like data glove devices and color caps to develop a piece of complex information about gestures provided by the user/human. Similarly, the virtual mouse that we'll discuss in this project is formed from OpenCV and PyAutoGUI.
>OpenCV is a python library for computer vision that is employed for capturing images from a webcam and PyAutoGUI is a python library that's used for specifying keyboard and mouse operations.
>Technologies like Artificial Intelligence and Machine Learning have become booming technologies that help us easily to work and provide commands to systems with no human interactions. So here the model is additionally a neighborhood-color hose technology which may help us work with our systems with ease.
- Webcam : 640x480 pixels or above
- Hard disk : 500GB
- Processor : Dual core or greater
- Ram : 8GB or more
- Also it needs Python and all the Requirements packeges from Requriment.txt needed
-
The Front-end is designed using PyQt5 QtDesigner tool. PyQt5 is a cross-platform GUI toolkit, a set of python bindings for Qt v5. The GUI mainly consists of two modes i.e., Gesture based interaction and virtual mouse. In Gesture based interaction mode we are using jester model which is trained under RT3D_16F and twenty-BN jester dataset. It is necessary to adapt continuously, identifying when a gesture was performed and only then assign a label to the prediction. For that we used a window of the 16 most recent frames obtained, which updates itself with each new frame acquired.
-
In virtual mouse mode we have created an AI based Mouse controller. This is implemented using media pipe library, this library is a free and open-source framework under apache2.0 which is fully extensible and customizable and works across several platforms like android, iOS, desktop or cloud, web and Iot.
We are detecting the hand landmarks and then it is tracked and mouse operations is performed based on these hand landmarks. Mouse operations can be performed by using PyAutoGUI library and hand landmarks can be found out by using findHands() and findPosition() methods of HandTracking Module. Then we are finding the tip of the all the fingers and checks which fingers are up. If only the index finger is up then moving mode is activated.
Similarly, if both index and middle fingers are up then left click operation is performed, only when the distance between these two-tip id’s is less than 40. PyAutoGUI.leftClick() method is used to perform left click operation of the mouse.
| Name | |
|---|---|
| Rithesh | ritheshrai5@gmail.com |
| Rohan J billava | rjbillava24091999@gmail.com |
| Adarsha k | adarshakanthabailu@gmail.com |
| Sebastian | sebastianthomas075@gmail.com |