PowerVision

A piece of software that reads a schematic, turns it into a SPICE netlist, and simulates it.

Currently this software is optimized for planar circuits that do not have transformers.

Requirements:

Virtual Environment:

 pip install virtualenv
 python -m venv powervision

 * To activate venv:
 source powervision/Scripts/activate

 * To deactivate venv:
 deactivate

TensorFlow:

 // If your computer has an NVIDIA GPU:
 pip install tensorflow[and-cuda]

 // If not:
 pip install tensorflow

OpenCV and other important libraries:

 pip install opencv-python numpy matplotlib
 pip install scikit-image
 pip install imutils

LTSpice:

 https://www.analog.com/en/resources/design-tools-and-calculators/ltspice-simulator.html

Clone this repo.
Create a Virtual Environment (venv) called powervision by following the steps in the previous section. This is important as this will isolate this software from the rest of your computer in case there are any clashes between libraries, and on Windows machines it allows you to install Tensorflow without facing the "long path" issue.
Install all required libraries and dependencies.
Upload the circuit you want to analyze, making sure that it is in the same directory as your powervision venv folder.
Run powervision.py using:
```
 python powervision.py
```
Follow the instructions, try it out with the provided examples in the example folder.

Uses image processing libraries and OpenCV on an image of a schematic in order to narrow down the search for components.
Crop the components found in the schematic and note their locations
Run images a trained CNN to identify the type of component.
Take each cropped image and do some more image processing to determine the orientation of the components
Generate a netlist matrix.
Run the matrix through a netlist generator, which also automatically adds gate drivers for MOSFETs
Calls LTSpice on the netlist, reads the RAW file and plots the data using matplotlib.