COMP 9517 WK7
2021 T1
Motion & Tracking
Applications
Scenarios
Image Subtraction
• Derive a background image from a set of video frames at the beginning of the video sequence
• Subtract the background image from each subsequent frame to create a difference image
• Threshold and enhance the difference image to fuse neighbouring regions and remove noise
Image subtraction algorithm
Sparse Motion Estimation
Sparse Motion Estimation
Detect Interesting Points
Sparse Motion Estimation
Similarity Measures
Similarity Measures
Deep Learning
Deep Learning
Traditional Approach vs DL
Traditional Approach vs DL
Neural Networks
NN: Perceptron
Neural Network Architecture
Why CNNs?
So what is different?
So what is different?
CNN Architecture
CNN Architecture
CNN: Convolutional Layer
CNN: Convolutional Layer
CNN: Convolutional Layer
CNN: Convolutional Layer
CNN: Convolutional Layer
CNN: Convolutional Layer
CNN: Pooling Layer
CNN: Pooling Layer
CNN: ReLU Layer
CNN: Dropout Layer
CNN: Dropout Layer
CNN: Dropout Layer
CNN: Training
CNN: Training
CNN: Training