Download Preliminary Project List

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

Document related concepts

Navier–Stokes equations wikipedia , lookup

Fluid dynamics wikipedia , lookup

Derivation of the Navier–Stokes equations wikipedia , lookup

Computational fluid dynamics wikipedia , lookup

Transcript
Fall Workshop Final Project Suggestions
Face Detection
Difficulty: Hard
Synopsis: Being able to identify a person from an image has obvious
applications, hence why face detection has become a important tool of the
surveillance state. The project consists of the first stage of facial recognition,
which is of detection of facial features. The larger goal of being able to
recognize a person from their features is beyond the short scope of this
workshop.
Inputs: A set of images of faces of various people.
Desired Outputs: Identification the regions of the image corresponding to
the eyes, nose and mouth of the person in the image.
Handwriting Detection
Difficulty: Easy
Synopsis: In this project you will build a system to detect handwritten
numbers and letters. There are several ways to create a program that
recognizes handwriting any you may choose the implement whichever you
want. One of the most common though is to use a nearest neighbors
algorithm and to train the algorithm on a large data set of presorted data.
Inputs: 50 pictures of hand written numbers
Desired Outputs: A list of numbers.
Heat Equation Simulation
Difficulty: Easy
Synopsis: In this project you will implement a numerical approximation of
heat dissipation through a surface (or volume). Using the heat equation and
numerical approximation you will create a program that can take in any
initial heat signature on a plane or 3d volume and iteratively compute the
heat flow in a time series.
Inputs: Initial temperature configuration and the thermal conductivity of the
2d material
Desired Outputs: A time series or movie depicting the heat flow in time.
Wave Equation Simulation
Difficulty: Easy
Synopsis: Simulate the 1, 2 or 3 dimensional wave equation by both using
analytic results and numerical approximation.
Inputs: An initial impulse or wave configuration
Desired Outputs: A time series of images or a movie of the wave
propagation.
Non-Linear Dispersion Simulation
Difficulty: Medium
Synopsis:
Inputs:
Desired Outputs:
Fluid Dynamics Simulation
Difficulty: Hard
Synopsis: Fluid dynamics constitutes a “hard” problem for mathematics
since the differential equations thought to govern it can be easily written
down but as of yet have no exact solutions. However, from the differential
equations alone the behavior of fluids can be approximated and modeled
extremely well. In this project you will use the shallow water equations to
simulate sloshing of the surface of a fluid given and certain initial
configuration.
Inputs: A 2d matrix giving an initial height map for a surface of a liquid.
Desired Outputs: A video stream simulating the relaxing a sloshing of the
liquid given the initial condition.
Motion De-Blurring
Difficulty: Medium
Synopsis:
Inputs:
Desired Outputs:
Object Tracking
Difficulty: Medium
Synopsis: performing an analysis on a stream of video to flag the continuity
of features between frames has many applications, from image stabilization
to security camera tracking to programing a webcam to watch your
experiment for you and text you when the mouse has reached the cheese.
Object Tracking: Stabilization
This project focuses either taking in several seconds of shaky video and
performing cropping and rotations to align the frames.
Inputs: A shaky video stream of a still image/still life
Desired Outputs: A video stream of the same still life with the shaking
removed.
Object Tracking: Movement Tracking
In this project, a stream of stable video with a moving object, say a thrown
ball, will be analyzed. The moving objects should be highlighted and
tracked across time.
Inputs: A video stream with fixed background and moving foreground
Desired Outputs: A video stream with moving objects highlighted
Edge/Line/Corner Detection
Difficulty: Easy
Synopsis: Edge detection is one of the simplest but most important
techniques in any kind of image analysis (I’m actually serious, I don’t think
this would be considered controversial by anyone). Edge detection is the
main tool that allows you to single out objects in a picture and separate
features from background data. In this project you will implement at least
one of the common edge detection algorithms and
Inputs:
Desired Outputs:
Stereo Image Analysis
Difficulty: Medium
Synopsis:
Inputs:
Desired Outputs:
Feature Classification
Difficulty: Medium/Hard/Research Level
Synopsis: The human eye takes in photons and transforms the information
into nerve impulses. These nerve impulses seem to be optimized to detect
different kinds of feature natural to our environment, in optical neuroscience
one of the tools used to suss out which of these features our brain seems
to prefer and which are left out and “assumed in postprocessing.” One of
the current leading theories is that our eyes would be evolutionary adapted
to encode image data from “natural images.” The point of this project would
be to determine what that means, by finding an algorithm that can sort 100
pictures into “natural images” (like forests, grass, etc.) and “synthetic”
images (pure noise, one tone, checker board, etc)
Inputs: 50 pictures
Desired Outputs: A classification of each picture as either “natural” or
“unnatural.”