Physical AI

59 Lessons( 20 week )

• Basics of Image Processing
• 1. Digital Image Representation (Pixels, Color Spaces, Histograms)
• 2. Filters: Blurring, Sharpening, Edge Detection (Sobel, Canny)
• 3. Morphological Operations: Dilation, Erosion, Opening/Closing
• 4. Libraries: OpenCV, Pillow
• Image Acquisition & Devices
• 1. Camera Types: RGB, IR, Depth Cameras, Stereo Vision, Industrial Cameras
• 2. Lens Selection, Lighting Conditions, Shutter Speed, Frame Rate
• 3. Using USB And CSI Cameras with Raspberry Pi or Jetson
• 4. Camera Calibration and Distortion Correction
• Image Preprocessing Techniques
• 1. Thresholding, Adaptive Thresholding
• 2. ROI Extraction, Resizing, Normalization, Augmentation
• 3. Feature Extraction: HOG, SIFT, ORB
• 4. Geometric Transformations: Perspective Correction, Affine Warp
• Feature Detection & Object Tracking
• 1. Background Subtraction, Motion Detection
• 2. Object Tracking: Centroid Tracking, Kalman Filters
• 3. Template Matching, Correlation Filters
• 4. Color Histogram Tracking
• Classical ML for Vision
• 1. ML Workflows with Scikit-Learn
• 2. Features vs Labels, Training/Testing Datasets
• 3. SVM, KNN, Decision Trees for Classification
• 4. Image Classification Using Classical Pipelines
• Visual Programming with Orange
• 1. Intro to Orange Data Mining for Image Workflows
• 2. Drag-And-Drop Pipelines for Feature Selection, Classification and Visualization
• 3. Custom Widgets for Image Inputs and Preprocessing
• 4. Integration with Scikit-Learn Models
• Deep Learning Basics
• 1. Artificial Neural Networks Recap
• 2. Forward and Backward Propagation
• 3. Introduction to TensorFlow/Keras and PyTorch
• 4. Setting Up GPU Environments (Colab/Jetson Nano)
• Convolutional Neural Networks
• 1. CNN Architecture and Layers
• 2. Pooling, Activation, Dropout, Batch Normalization
• 3. Image Classification Using CNNs (MNIST, CIFAR-10)
• 4. Transfer Learning (VGG, Resnet, MobileNet)
• Object Detection Models
• 1. YOLO (You Only Look Once): YOLOv5, YOLO-NAS
• 2. SSD, Faster-RCNN: Pros And Cons
• 3. Real-Time Detection using Webcam/Video Stream
• 4. Annotating Custom Datasets (RoboFlow, CVAT)
• Industrial Computer Vision Applications
• 1. Visual Inspection: Surface Defects, PCB Inspection, Misalignment
• 2. OCR And Barcode/QR Detection
• 3. Pose Estimation and Measurement
• 4. Object Counting and Sorting
• Depth & 3D Vision
• 1. Stereo Vision, Depth Sensors (RealSense, ZED)
• 2. Point Cloud Basics and 3D Reconstruction
• 3. SLAM Basics for Spatial Mapping
• 4. Open3D And PCL Usage
• Image Segmentation
• 1. Semantic vs Instance Segmentation
• 2. UNet and Mask-RCNN Architectures
• 3. Dataset Preparation and Annotation for Segmentation
• 4. Applications: Robotic Grasping, Crop Segmentation
• Hardware for Edge Vision
• 1. Jetson Nano/Xavier, Coral TPU, OpenVINO Devices
• 2. Power, Cooling, Camera Interfaces, Deployment Options
• 3. Accelerating Inference with TensorRT, ONNX, TFLite
• Edge Deployment Pipelines
• 1. Building Docker Containers for Models
• 2. Flask/FastAPI REST APIs for Camera Input
• 3. Integrating Camera + AI Models with Node-RED
• 4. Offline vs Connected Deployment Considerations
• Real-time Video Analytics
• 1. Frame Skipping, Threading for Performance
• 2. Handling Low-Latency Inference Pipelines
• 3. Stream Processing and MQTT Integration
• 4. Alerting On Anomaly or Defect Detection