Versions Compared

Old Version 53

changes.mady.by.user Sanja Bonic

Saved on

compared with

New Version 54

changes.mady.by.user Sanja Bonic

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...


Draw Simplified Icon Representations of Nouns Given via Various Inputs

As part of a larger ML project that will be open sourced, this part of the project is about drawing several icon representations of nouns within one image. You will take as input a list of nouns and let the computer represent those nouns via icons/sketches in an image file.


First steps:

  • Decide the possible input sources (CLI, web, voice, ?)
  • Research useful libraries and decide how to do error handling (receive full sentences and extract just nouns via entity recognition or be sure to receive only nouns)
  • Decide whether to go for sketches or icons (various approaches and libraries available open source)
  • Decide how to combine representations into an image (positioning)


This project helps you to:

  • Get started or advance with AI/ML topics
  • Get creative
  • Contribute to open source software publicly and visibly on your GitHub profile
  • Get into research (writing papers, learning how to publish)


Prerequisites:

  • Solid programming skills
  • Git


Nice to have:

  • Rudimentary or advanced knowledge of AI/ML (not a requirement, but a bonus for yourself to get started quicker)


This topic is available at all levels.
For more information, please contactstefan_schmid@univie.ac.at or sanja.bonic@univie.ac.at.



Determine

Evaluate Composition

Value

in Images

As part of a larger ML project that will be open sourced, this part of the project is about determining evaluating the composition value in of images. You will find images with a license that allows usage in research projects and find a way to determine their composition value. You will then feed this information to an ML algorithm that will be used in the subsequent parts of this overall project.


First steps:

  • Locate > 30,000 images with a suitable license
  • Research existing libraries for this topic and write a report that is part of your project
  • Decide on the classification scale you want to use (e.g. good/neutral/bad, 1-10, ?)
  • Classify the images


This project helps you to:

  • Get started or advance with AI/ML topics
  • Get creative
  • Contribute to open source software publicly and visibly on your GitHub profile
  • Get into research (writing papers, learning how to publish)


Prerequisites:

  • Solid programming skills
  • Git


Nice to have:

  • Rudimentary or advanced knowledge of AI/ML (not a requirement, but a bonus for yourself to get started quicker)
  • Photography knowledge


Reading starters:

  • Marcelo Siqueira, Longin Jan Latecki, and Jean Gallier "Making 3D binary digital images well-composed", Proc. SPIE 5675, Vision Geometry XIII, (17 January 2005); https://doi.org/10.1117/12.596447
  • Ngo, Phuc & Passat, Nicolas & Kenmochi, Yukiko & Talbot, Hugues. (2013). Well-composed images and rigid transformations. 2013 IEEE International Conference on Image Processing, ICIP 2013 - Proceedings. 3035-3039. 10.1109/ICIP.2013.6738625.
This topic is available at all levels.
For more information, please contactstefan_schmid@univie.ac.at or sanja.bonic@univie.ac.at.



Determine Position and

Evaluate

Calculate Area Percentage in Images

As part of a larger ML project that will be open sourced, this part of the project is about recognizing the position of objects in images and defining calculating the area percentage these objects take within the image. You will then feed this information to an ML algorithm that will be used in the subsequent parts of this overall project.


First steps:

  • Locate > 30,000 images with a suitable license
  • Research existing libraries for this topic and write a report that is part of your project
  • Decide whether you want to work with artificial images (e.g. banners used on company websites, blogs, magazines) or natural pictures (pictures where the original source can be assumed to be natural)


Research questions:

  • Which parts of a picture form a single object?
  • Which percentage of the image is taken up by each individual object?


This project helps you to:

  • Get started or advance with AI/ML topics
  • Get creative
  • Contribute to open source software publicly and visibly on your GitHub profile
  • Get into research (writing papers, learning how to publish)


Prerequisites:

  • Solid programming skills
  • Git


Nice to have:

  • Rudimentary or advanced knowledge of AI/ML (not a requirement, but a bonus for yourself to get started quicker)
This topic is available at all levels.
For more information, please contactstefan_schmid@univie.ac.at or sanja.bonic@univie.ac.at.

...