ImageEye: Batch Image Processing using Program Synthesis
summary
A synthesis pipeline that learns programs in a neuro-symbolic DSL uses Amazon recognition detectors and an enumerative synthesizer with equivalence-reduction and goal-directed pruning to auto-generate batch edits. Implemented as the tool ImageEye with a GUI, it solves 48 of 50 benchmark tasks (96%), times out on the two largest, has median synthesis ≈1s and mean ≈15s, and typically needs about four example images to converge.
AI-written summary · about
chapters (6)
- 0:00ImageEye introduction
- 3:13Neurosymbolic DSL
- 4:59Program synthesizer
- 8:48Demo of ImageEye graphical user interface
- 9:51Benchmarks evaluation
- 12:03Q&A
related talks
01h 18mComputer Vision after the Victory of Data
27mComputer vision for automated quality inspection
36mApps That See 👀 Practical AI Vision for Developers
49mGraph Based Processing of Big Images
01h 26mBuilding an Image Search Engine with Generative Computer Vision Models
45mBe My Eyes - Leveraging Generative AI to help the blind and visually impaired people
38mHow Artificial Intelligence has changed your life
25mPersonalizing Your Images with AI-Powered Features