Qualcomm is looking to explore the applicability of video diffusion models to various downstream applications, such as image editing or facial reenactment, by investigating novel architectures and training strategies to improve video generation quality and efficiency.
Requirements
- video diffusion models
- image editing
- facial reenactment
- novel architectures
- training strategies
- conditioning mechanisms
- baseline models
Responsibilities
- Investigate novel architectures and training strategies to improve video generation quality and efficiency.
- Explore conditioning mechanisms to extend existing video diffusion models to downstream applications.
- Generate or curate use-case specific datasets to support model development and evaluation.
- Implement and benchmark baseline models for comparative analysis, including thorough ablation studies.
- Optimize mod
Other
