Day 6: Computer Vision

What Is a Neural Network?

  • A computing system loosely inspired by the human brain
  • Made up of layers of connected nodes (neurons)
    • Input layer: receives raw data (numbers, pixels, text)
    • Hidden layers: find patterns in that data
    • Output layer: produces the final prediction
  • Each connection has a weight — how much does one neuron influence the next?
  • Stack enough layers and the network can learn almost anything

Neural Network Demo

Will I Do Well on My Test?

Hands-On

Try out the neural network yourself

Reflection

Did the neural network produce the results you expected?

Computer Vision is Everywhere

Computer Vision is Everywhere

Computer Vision is Everywhere

Computer Vision is Everywhere

How Do Computers See?

How Do Computers See?

  • Computers see images as grids of numbers — pixels
  • Each pixel is a number from 0–255 (or 3 numbers for color: R, G, B)
  • A 640×480 photo = nearly 1 million numbers
  • The challenge: how do we find meaning in all those numbers?

How Do Computers See?

  • Answer: A special type of neural network called a Convolutional Neural Network or a CNN
    • Takes in an image
    • Detects edges, corners, and other shapes
    • Returns results on what it sees and where

How Do Computers See?

  • Many types of computer vision algorithms:
    • Face Detection (is there a face in the picture?)
    • Object Detection (what is in this image and where?)
    • Pose Estimation (where are the body parts?)
    • Gesture Recognition (what hand gesture is this?)
    • Image Segmentation (what pixel belongs to what object?)

Face Detection

  • Question: “Is there a face in this image, and where is it?”
  • Detects faces and draws bounding boxes around each one
  • Not facial recognition!
  • Used in:
    • Camera autofocus algorithms
    • Security cameras detecting people in the scene
    • Counting people attending an event

Demo

Face Detection Notebook

Hands-On

Face Detection Notebook

Object Detection

Object Detection

  • Question: “What is in this image, and where is it?”
  • Detects multiple objects and draws bounding boxes around each one
  • Examples:
    • Self-driving cars spotting pedestrians and vehicles in real time
    • Counting inventory in a warehouse automatically
    • Accessibility tool for people with vision impairments

Demo

Object Detection Notebook

Hands-On

Object Detection Notebook

Pose Estimation

  • Question: “Where are the body parts?”
  • Detects key landmarks on a person — joints, hands, face points
  • Examples:
    • Fitness apps checking your squat form
    • Video games that respond to your movement (no controller needed)
    • Physical therapy tools tracking patient recovery

Demo

Pose Estimation Notebook

Hands-On

Pose Estimation Notebook

Fun With Poses

Fun With Poses

  • How about if didn’t show the camera view?
  • And instead, just plotted the body joint positions…
  • In a 3D Scene?

Demo

3D Pose Avatar Notebook

Hands-On

3D Pose Avatar Notebook

Tomorrow

  • Continue learning about computer vision
  • Gesture recognition and Image Segmentation
  • Two other mini-apps