Projects per year
CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by finetuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN’s performance is the highest yet, it remains less than 60% AP, suggesting further work is needed for the cross-depiction problem.
|Title of host publication||Computer Vision - ECCV 2016 Workshops. ECCV 2016.|
|Editors||Gang Hua, Hervé Jégou|
|Number of pages||17|
|Publication status||Published - 18 Sep 2016|
|Event||14th European Conference on Computer Vision, ECCV 2016 - Amsterdam, Netherlands|
Duration: 8 Oct 2016 → 16 Oct 2016
|Name||Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)|
|Conference||14th European Conference on Computer Vision, ECCV 2016|
|Period||8/10/16 → 16/10/16|
- Cross-depiction problem
- Object recognition
ASJC Scopus subject areas
- Theoretical Computer Science
- Computer Science(all)
FingerprintDive into the research topics of 'Detecting people in artwork with CNNs'. Together they form a unique fingerprint.
- 1 Finished
Cosker, D., Bilzon, J., Campbell, N., Cazzola, D., Colyer, S., Fincham Haines, T., Hall, P., Kim, K. I., Lutteroth, C., McGuigan, P., O'Neill, E., Richardt, C., Salo, A., Seminati, E., Tabor, A. & Yang, Y.
1/09/15 → 28/02/21
Project: Research council