Computer Vision on Raspberry Pi - Beginner to Advanced

Format the SD card, download and unzip NOOBS 2.4.0, then select all files and drag them to the SD card drive to install the NOOBS operating system for Raspberry Pi.

Learn how to flash the NOOBS operating system to an SD card, boot the Raspberry Pi, and install Raspbian with a Pixel welcome screen, using HDMI, keyboard, and power setup.

Set up a Raspberry Pi 3 with NOOBS, connect to the internet, and install Python and OpenCV for image processing, then tackle five projects like face and car detection.

Learn how to use the Python print function to print messages and variable values on a Raspberry Pi running noobs, including simple arithmetic like 10 minus 3.

Explore decision making in Python through conditional statements, input handling with raw_input, and if/elif/else logic, demonstrated via a coffee and tea choice workflow on Raspberry Pi.

Explore for loop fundamentals on Raspberry Pi, initialize and iterate over numeric, string, and mixed lists, print results, and learn append usage and range-based looping.

Learn to define and call functions in Python using def, with four parameters (English, German, Spanish, French), and explore four value-passing methods and formatted printing.

Explore Python objects by creating variable types like int, list, tuple, and dictionary, and define a function; print their types to see int, list, tuple, dictionary, and function.

Explore Python modules by building a vehicle class with attributes like tyres and seats, instantiate objects with __init__, and print values to see how changes affect output.

Develop and test image workflows on Raspberry Pi by loading, displaying, and saving images with Python and OpenCV, using cv2.imread, cv2.imshow, and cv2.waitKey to manage image visuals.

Discover how to scale images with cv2 by resizing with cv2.resize using cubic, area, and linear interpolation, including factor-based and dimension-based options, while reading and displaying the image.

Flip images using OpenCV in Python by reading image_3.jpg, displaying originals, and applying cv2.flip with 1 for horizontal, 0 for vertical, and -1 for both.

Vary the brightness of images using add and subtract operations with NumPy and OpenCV. Read the image, create a brightness matrix with np.ones, and display original, added, and subtracted results.

Blurring and sharpening images on a raspberry pi using OpenCV: blur with a 9x9 normalized box filter and sharpen with a 3x3 kernel.

Apply binary thresholding to a grayscale image using cv2, converting color images to gray (BGR2GRAY) and displaying the binary result from coins example.

Perform image segmentation with contours by applying canny edge detection to reveal boundaries. Extract and draw all contours on the image using findContours and drawContours.

Capture real-time human faces from a webcam, detect and crop them using haarcascade_frontalface_default.xml, and save the 250x250 crops to the human_faces folder.

Explore how artificial neural networks imitate biological neurons by processing multiple inputs through synapses with weights, applying activation functions to produce continuous, binary, or categorical outputs.

Study activation functions in neural network, including sigmoid (logistic), tanh, ReLU, leaky ReLU, and softmax. Learn how weighted sums trigger neuron outputs and produce probabilities.

Explore how neural networks process inputs through input, hidden, and output layers, learn via backpropagation, and refine weights with activation functions to predict outcomes and minimize error using TensorFlow.

Explore neural network optimization with gradient descent variants: stochastic, batch, and mini-batch, learning to minimize error via forward and backward propagation across deep networks.

Demonstrates implementing simple neural networks in Google Colab with Python, using weights and inputs, NumPy arrays, and dot products for single and multi-input predictions, including hidden layers.

Explore gradient descent and hot and cold learning methods to iteratively adjust weights using learning rate, to minimize squared error and approach the ground truth prediction.

Learn how convolutional neural networks classify images, detect objects, and segment regions using kernels, ReLU activation, max pooling, flattening, and fully connected layers.

Combine convolutional layers, rectified linear activation, max pooling, and fully connected stages to classify images using a cnn. Explain backpropagation, gradient descent, and hyperparameters that minimize digit error across epochs.

Learn to build and train a convolutional neural network on MNIST with TensorFlow and Keras, covering one-hot encoding, data normalization, and model evaluation on train and test sets.