State of the art in AI
Taken from this excellent course: Deep Learning Lectures
Speech to text
Waveform ā Text
Computer vision
Image ā class
Image ā (class, bounding box)
Image ā (class, shape)
Image ā (class, bounding box)
Image ā (class, shape)
Image ā (class, shape)
Image ā facial landmarks
Image ā facial landmarks
Natural language processing (NLP)
Text ā text (different language)
Text ā syntax tree
Text ā syntax tree
Text ā text (probable short answer)
Text ā text (query)
Text ā text (query)
Computer vision & NLP
Image & text (question) ā text (answer)
Image ā text (description)
Image ā text (description)
Image translation
Image ā image (with artifacts)
Image ā image styled as the other
Image ā Image (higher resolution)
Image ā image styled as the other
Image ā Image (higher resolution)
Audio generation
Waveform ā Waveform (continued)
Guess which one is generated ?
Image generation
Vector (random) ā image
Text ā image
Science - Genomics, biology, chemistry, physics
DNA sequence ā drug
Protein sequence ā folding shape (protein properties)
Chemical structure ā properties
Incompressible Euler equations (Navier-stokes for fluids) ā guess of the solution
100x speedup in solving time
Gaming
Image sequence ā next action
More models and tasks
Papers with Code - Browse the State-of-the-Art in Machine Learning