Agriculture and Environment Natural Resources

Area/s: Agriculture and Environment | Natural Resources

Organization: Universitat Autònoma de Barcelona

Research theme code : RLA-UAB-03

UAB invites applications for a postdoctoral fellowship focused on machine learning based data compression. Fellows will work on developing AI models for compressing data coming from different sources (e.g., remote sensing, synchrotron data, …), for different scenarios (on-board satellites, on-the-ground), and contemplating lossy, lossless and near-lossless compression. Ideal candidates will have a background in AI and eager to drive innovations with real-world industrial applications.
Our work aims to address the challenges posed by the ever-growing volume of data generated daily. Here are some illustrative examples:

  • Hyperspectral Coding: Multispectral and hyperspectral images contain numerous spectral bands, each selectively detecting different frequencies of light. These bands provide valuable information for Earth and space observation tasks. However, their high dimensionality presents compression challenges.
  • Satellite Constellations: With the rise of smaller, more affordable satellites, constellations are becoming common, but data transmission from satellites to Earth has restricted band-with capabilities.

Some benefits due to Machine Learning based approaches are as follows:

  • Improved Compression Ratios: Machine learning algorithms can exploit redundancy across spectral bands, leading to more compact representations.
  • Energy Efficiency: ML-based approaches can reduce energy consumption during compression, and lightweight approaches can be devised.
  • Guaranteed Image Quality: By optimizing for specific purposes (e.g., land cover classification, synchrotron data exploitation), ML models can maintain desired image quality.

The selected fellow would be integrated in the Department of Information and Communications Engineering at the UAB, within the research group on Interactive Coding of Images (GICI). The selected fellow would collaborate with several partnering institutions around the world (including space agencies, synchrotron premises, private companies, …).
Fellows can also suggest some other application domain, as long as it is related to data compression.

Key Responsibilities:

Design and develop AI models for data compression.

  • Design and develop AI models for data compression.
  • Research and design algorithms for real-world applications.
  • Supervision of PhD students.

Qualifications:

  • PhD in Computer Science, Data Science, Mathematics or related fields.
  • Strong knowledge of AI techniques, including deep learning, image processing.
  • Interest in interdisciplinary collaboration at the intersection of technology and real-world applications.
  • Strong collaborative skills.

Principal Investigator: