Export
Dena Bazazian
· 7 previous affiliations
- University of PlymouthSince 2022
- Assistant Professor · Since 2022
- University of Bristol2020–2022
- Postdoc · 2020–2022
- Researcher · 2019–2020
- Universitat Autònoma de Barcelona2015–2019
- Postdoc · 2018–2019
- PhD student · 2015–2018
- NAVER Cloud (South Korea)2018–2019
- Researcher · 2018–2019
- Researcher · 2017
- Universitat Politècnica de Catalunya2012–2015
- Researcher · 2012–2015
Research Overview
Dena Bazazian focuses on computer vision, specifically addressing unconstrained word spotting in natural scene images and advancing diversity within the machine learning community. Her early trajectory emphasized text detection using fully convolutional networks and character recognition techniques. More recently, her associated research themes indicate an interest in generative image composition and latent space alignment.
Research Trajectory
The available evidence for Dena Bazazian's research trajectory is concentrated in 2018, highlighting both technical contributions to computer vision and active efforts to improve inclusion within the academic community.
In her technical work, Bazazian led research on text detection in natural environments. As the first author of Word Spotting in Scene Images Based on Character Recognition, she addressed the challenge of unconstrained word spotting. This work utilized a Fully Convolutional Network to generate heatmaps for character classes, combining these with a text proposals approach and a rectangle classifier to identify query words in scene images.
Alongside her technical research, Bazazian has been a prominent organizer and advocate for diversity in the field. She was a co-author and co-organizer for major community initiatives, including WiCV 2018: The Fourth Women in Computer Vision Workshop at CVPR and WiCV at ECCV2018: The Fifth Women in Computer Vision Workshop. These publications document structural efforts to increase the visibility of female researchers, provide mentorship opportunities, and address gender-related biases in computer vision and machine learning.
Current Directions
While the specific publications in the provided corpus date to 2018, Bazazian's broader research profile includes more recent themes from 2022 centered on advanced generative AI and image synthesis. Her recent thematic associations point toward the research direction of Resolving Spatial Misalignment in Conditional Image Synthesis and the area of Semantic Inheritance through Latent Space Alignment.
Additionally, these recent themes indicate an interest in the practical challenges of image editing and generation. This includes work in the broader area of Seamless Image Composition in the Gradient Domain, which addresses the visual discontinuities or "seams" that occur when editing images or pasting synthetic objects. Related thematic interests also encompass the research direction of Photometric Consistency in Generative Object Insertion and the area of Few-Shot Adaptation of Generative Models, focusing on adapting pre-trained generative models to new domains with minimal data.
Profile updated
| Title | Venue | Year |
|---|---|---|
| WiCV 2018: The Fourth Women in Computer Vision Workshop | Workshops @ CVPR | 2018 |
| WiCV at ECCV2018: The Fifth Women in Computer Vision Workshop | Workshops @ ECCV | 2018 |
| Word Spotting in Scene Images Based on Character Recognition | Workshops @ CVPR | 2018 |
Loading new papers…
Loading research neighbors…
See something wrong? Report a mistake