Petru-Daniel Tudosiu
@DTudosiu
Followers
9
Following
19
Media
4
Statuses
14
Generative Modelling Research Scientist @HuaweiUK | Previously @Kingsimaging | Opinions are my own and not the views of my employer
Joined March 2023
Research Scientist Internship - https://t.co/uyCAHiyR37 Research Engineering Internship - https://t.co/mn6E6kRGo5 Research Scientist Contractor - https://t.co/U7hHCyBC1C 9/9
1
0
1
If you are interested in pushing the boundaries of Vision Language Modelling and Interleaved Text-to-Image please do not hesitate to get in touch or apply directly below. 8/9
1
0
1
With MuLAn, we aim to encourage the development of novel generation and editing technology, in particular layer-wise solutions. 7/9
1
0
1
The pipeline is composed of three modules: image decomposition for instance discovery and extraction, instance completion to reconstruct occluded areas, and image re-assembly. 6/9
1
0
1
To build MuLAn, we developed a pipeline based on pretrained-general-purpose models which decomposes a monocular RGB image into a stack of RGBA layers comprising of background and isolated instances. 5/9
1
0
1
To facilitate generative research using multi-layer representation, we release MuLAn, a new dataset which provides multi-layer RGBA decomposition of natural images. 4/9
1
0
1
We believe this is to due them failing to take advantage of the natural instance-level compositionality of scenes due to the typically flat nature of rasterized RGB output images. 3/9
1
0
1
Even with recent T2I progress having precise controllability and good prompt comprehension remains challenging. Current approaches are cumbersome and require substantial amount of manual work. 2/9
1
0
1
My first Huawei publication has been accepted at CVPR 2024! We have released MuLAn a multi-layer annotated dataset for controllable text-to-image generation. Paper: https://t.co/HfdQXtJsqP Homepage: https://t.co/fJ2rSWIZQA Dataset: https://t.co/xpLVXeVSVF 1/9
huggingface.co
1
0
2
Even with recent T2I progress having precise controllability and good prompt comprehension remains challenging. Current approaches are cumbersome and require substantial amount of manual work. 2/9
0
0
0
The resulting work will be likely submitted at a top tier conference (NeurIPS, ICLR, ECCV, ICCV, etc) as well as for patenting . Please get in touch with either myself or Yiming Fu ( https://t.co/fMHu9UxUHn) to submit your application.
linkedin.com
Experience: Faraday Future · Education: 澳大利亚蒙纳士大学 · Location: Los Angeles Metropolitan Area · 500+ connections on LinkedIn. View Yiming Fu’s profile on LinkedIn, a professional community of 1 billion...
0
0
0
A new Research Engineering Internship position has been opened ( https://t.co/y04HM3bpnp). The research engineering project aims to improve an existing pipeline in terms of speed, performance and reliability, and add human-in-the-loop capabilities.
1
1
2
Good news in medical generative AI!🎉🎉 We just published the preprint of the MONAI Generative Models Extension! Check out our latest experiments with 2D and 3D data, ControlNets and 3D Cascaded Diffusion Models! https://t.co/CRcgi17PZB
#AI #MedicalImaging #GenerativeModels
4
55
187
Happy to announce that our work led by @marksgraham_ on diffusion models applyed to out-of-distribution was accepted at CVPR workshop on Visual Anomaly Detection. This work was done during my PhD at @KingsImaging under @mjorgecardoso.
Our work applying diffusion models to out-of-distribution detection has been accepted at the CVPR workshop on Visual Anomaly Detection. Paper: https://t.co/y9kKqsLOlJ Code: https://t.co/32dpwUeIhz with @mjorgecardoso, @Warvito, @PTudosiu, Seb Ourselin, Parashkev Nachev
0
0
1