Goal 1: Research & Development
Our vision is that everything Visual should be Describable and Searchable
We aim to develop the next generation of computer vision algorithms that are able to analyse, describe and search image and video content with human-like capabilities and far beyond. Computer vision has made tremendous progress in recognition over the last decade. It is now possible to search and recognize object instances – such as paintings, book covers and buildings using a mobile phone (e.g. using the Google Goggles app) – and to organise and search images by topic or content, by recognising thousands of object categories in images and videos (such as vehicles and animals – see above). Yet, despite this success, the technology is still significantly more primitive than vision in humans and cannot satisfactorily address tasks more advanced than recognizing and localizing entire objects. Humans are able to describe scenes and video activities that they have never seen before; search for objects and events given only a description of a handful of words; and analyse (count, localize, delineate) a vast array of disparate visual sources. With Big Data such capabilities must also be scalable to billions of images and videos.
Seebibyte Project research will focus in particular on two key areas: 1) Visual Description and Analysis; and 2) Visual Search and Distillation.
Goal 2. Transfer & Translation
We aim to transfer the latest computer vision methods into other disciplines and industry.
This goal has two aspects. The first is to apply the new computer vision methodologies to ‘non-natural’ sensors and devices, such as ultrasound imaging and X-ray, which have different characteristics (noise, dimension, invariances) to the standard RGB channels of data captured by ‘natural’ cameras (iphones, TV cameras). The second is to seek impact in a variety of other disciplines and industry which today greatly under-utilise the power of the latest computer vision ideas.
We want the software developed in this project to be taken up and used widely by people working in industry and other academic disciplines. As the project progresses, we will release new open source software, datasets, and demonstrators on the project website. We will also be organising national and university-wide outreach activities – including “Show and Tell” events, where we will demonstrate software developed by our researchers.