COSC428 Computer Vision

Flashcards for Computer Vision Review

What is the goal of computer vision?

Recognizing objects and their motion.

Vision is

Inferential

Name some fields related to computer vision.

Artificial Intelligence, Automatic Control, Robotics, Computational Intelligence, Robot Vision, Machine Learning, Cognitive Vision, Computer Vision, Machine Vision, Signal Processing, Non-linear SP, Multi-variable SP, Physics, Optics, Image Processing, Statistics, Geometry, Optimization, Smart Cameras, Biological Vision, Mathematics, Neurobiology, Imaging

What is the first step to Intelligence and Perception?

First to understand how we perceive the world then to teach the machine to interpret the world based on primitive data it has received

Name the Perceptual modalities

Tactile, Gustatory, Visual, Auditory, Olfactory

What is the sensory gap in vision computing?

The gap between the object in the world and the information in a (computational) description derived from a recording of that scene.

What is the goal of the shape-from-contour module?

Derive information about the orientation of the various different faces.

What is the semantic gap in vision computing?

The lack of coincidence between the information that one can extract from the visual data and the interpretation that the same data have for a user in a given situation.

Name some cues available in the visual stimulus to recover 3D information.

Motion, Stereo, Texture, Shading, Contour, Time of flight

Define grey images.

Image Intensities (brightness) are discretely sampled, and the sampled values are quantized to a discrete set of values (integers).

What does grey level varies from?

From 0 (black) to 255 (max brightness, maximum response the eye can make)

What is a pixel?

Pixel – receptor in the retina

Why is difficult in vision computing - taking the human visual system for granted?

The processing capability of human visual systems is often taken for granted

Vision is inferential:

Light

Vision is inferential:

Prior Knowledge

Computer Vision = _

Inference → Computation

Boundary Detection: _

Local cues

Computer Vision Processing _

Simulate human image perception

Computer Vision Pre-processing: _

Noise removal, contrast enhancement etc.

Computer Vision Early processing: _

Find useful info from raw images

Computer Vision Late Processing: _

Find objects and meanings from the useful info

What are examples of Low level image processing?

Image compression, Noise reduction, Edge extraction, Contrast enhancement, Segmentation, Thresholding, Morphology, Image restoration

Why is Vision Interesting in Psychology?

~ 50% of cerebral cortex is for vision. Vision is how we experience the world.

Why is Vision Interesting in Engineering?

Want machines to interact with world. Digital images are everywhere.

Recognition - Shading: _

Lighting affects appearance

List few Vision Based Interfaces

Hand tracking, Hand gestures, Arm gestures, Body tracking, Activity analysis, Head tracking, Gaze tracking, Lip reading, Face recognition, Facial expression

Vision systems - passive includes _

laser scanner, structured light, time-of-flight, images

What are examples of Depth Imaging Technologies?

Active Stereo, Structured Light, KINECT, Time Of Flight

The Image Classification Challenge (IMAGENET) includes _

1,000 object classes. 1,431,167 images

Smart Computer Vision needs _

Deep Learning

Approaches to Vision: Modeling + Algorithms _

Build a simple model of the world and Find provably good algorithms.

Approaches to Vision: Engineering _

Focus on definite tasks with clear requirements and Try to build reusable modules

Related Fields to Vision : _

Graphics, Visual perception, Neuroscience, AI / machine learning, Math, Optimization

Computer Graphics needs _

Shape Model Reflectance Model Illumination Model

Inverse Modeling (a.k.a. “traditional” vision) needs _

geometry, physics computer algorithms real photos

Visual Modeling includes _

shape, light, motion, optics, images

Computer vision impacts graphics through _

image-based rendering, model acquisition, motion capture, perceptual user interfaces, special effects, image editing

Graphics impacts computer vision through _

reflectance, transparency, shape modeling

Deep Learning in Real-time - ~NZ$100 _

Neural Network Compute Stick from Movidius (Intel) has 100Gflops of computing power

MagicBook project - Collaboration with _

Gavin Bishop

Boss - 1st prize DARPA Urban Challenge - _

Carnegie Mellon Uni

Self Driving Cars - Now? _

Cars are already driving themselves on roads in California, Texas, Arizona, Washington, Pennsylvania, and Michigan but restricted to specific test areas and driving conditions.

predators can be recognised from a thermal camera by _

Deep learning using both shape and motion

NZ Scott Base, Antarctica is _

Accurately counting seals every 15 min with 187 Mpixel cameras

MBIE project: _

enable UAVs to use tools in complex dynamic environments

Trees can be recognised by _

Deep learning

volume for 3D surface reconstruction is done by _

Poisson surface reconstruction

Deep learning can autonomously generate semantic models to _

Recognise Trees, and terrain via deep learning

Very accurate point clouds can built within minutes through _

NeRF based deep learning

Vine pruning robot - _

PRAISE

Unreal Engine simulation - _

advanced underwater drone simulator for rapid prototyping of navigation and AI-based image recognition

Question - Name four different types of camera technologies for acquiring image depth values.

structured light camera, time-of-flight camera, stereo camera, LIDAR

For all depth cameras, _ can cause noisy depth values.

reflective (e.g. wet) surfaces

What are disadvantages of Structured light camera?

Cannot work in direct sunlight, Cannot work closer than 0.5m because the projected pattern of dots become too close together in the image, Cannot work further away than about 3.5m, Motion blur

What are disadvantages of Time of flight camera?

Cannot work in direct sunlight, Limited range due to low intensity infra-red light, Accuracy is independent of distance

What are advantages of Stereo camera?

Potential for highest resolution, Colour is also available for each pixel (as well as depth), Works well in direct sunlight. Works for motion (if well illuminated)

What are disadvantages of Stereo camera?

Noisy depth values in low ambient light, Accuracy decreases with distance, Many gaps in depth values in image regions without features

What are advantages of LIDAR camera?

Good range, Accuracy is independent of distance, Works well in direct sunlight

What are disadvantages of LIDAR camera?

Low resolution, Low frame rate, Has moving parts