Transfer and Translation of Computer Vision Research
The VisualAI programme grant started in Dec. 2020 with the following overall objective:
An open world Visual-Transformer (VT), a machine capable of ingesting visual and multi-modal data - including audio and text - and producing a robust, informative, and trustworthy interpretation of it. The VT will operate with human-like flexibility.
In this talk, I described how we pursued transfer and translation of computer vision research for the Seebibyte programme grant (2016 to 2020). I also talked about how we should pursue this goal for the new VisualAI programme grant (2021 to 2026). Here are the slides from the talk (odp, pdf).
Manual Annotation of Radiology Images using VGG Image Annotator (VIA)
We applied our VGG Image Search Engine (VISE) software to identify all the duplicate and transformed (e.g. rotation, resize, etc.) copies in the digital asset collection of the Ashmolean Museum containing 262007 images. In this presentation, we shared our results with the staff of the Ashmolean Museum on 04 Feb. 2020 (VGG, Oxford, UK)