Title:Fully Funded PhD in Recognition Regardless of Depiction
Posted: June 24, 2017
Company/Institution: University of Bath
Location: Bath, UK
Department: Computer Science
Description: Recognising objects in images is one of the most significant problems in all of Computer Vision. Contemporary methods using neural network are capable of truly impressive results, being able to positively identify a wide range of objects in a wide of conditions to accuracies exceeding 90%. Yet experimental evidence suggests that all systems for recognition are strongly biased to photographs - all of them struggle when input images include artwork.
This project builds on leading work at Bath that provides the only system able to recognise objects in both artwork and photographs [1,2,3].
The successful applicant will join the Visual Computing group at Bath. You will build a neural architecture specifically designed to generalise to recognise more objects in a far wider range of different types of artwork than is currently possible. You will be expected to continue its tradition of publishing in the highest quality forums, and to travel to conferences to present work to an international audience.
This work underpins many applications including content based search, the protection of visual intellectual property, and accessible computing. Other people will take the lead in developing such applications, but you will be expected to assist them with code, data, and conversation.
To learn more about this project, including a structured programme of study and it is value to industry contact the lead investigator Peter Hall on firstname.lastname@example.org.
 Q. Wu, H. Cai, P. Hall, Learning Graphs to Model Visual Objects across Different Depictive Styles. European Conference on Computer Vision, 2014
 H. Cai, Q. Wu, P. Hall, Beyond Photo Domain Object Recognition: Benchmarks for the Cross Depiction Problem. ICCV Workshop on Transferring and Adapting Knowledge in Computer Vision, 2016
 N. Westlake, H. Cai, and P. Hall. Detecting People in Artwork with CNNs. European Conference on Computer Vision. Springer International Publishing, 2016.