Advancements in Computer Vision and Image Generation

The field of computer vision and image generation is moving towards more efficient and effective methods for adapting models to various tasks and datasets. One of the key directions is the development of prompt-tuning frameworks that allow for flexible and lightweight adaptation of foundation models to downstream tasks. Another significant trend is the improvement of image generation models, including the development of methods for consolidating and unifying the capabilities of diverse models into a single one. Noteworthy papers in this area include:

  • Geometric Consistency Refinement for Single Image Novel View Synthesis via Test-Time Adaptation of Diffusion Models, which improves the geometric consistency of images generated by diffusion models.
  • DMM: Building a Versatile Image Generation Model via Distillation-Based Model Merging, which proposes a novel model merging paradigm for consolidating multiple models into a single versatile text-to-image model.
  • Learning Optimal Prompt Ensemble for Multi-source Visual Prompt Transfer, which introduces an adaptive framework for multi-source prompt transfer that learns optimal ensemble weights by jointly optimizing dual objectives.

Sources

Geometric Consistency Refinement for Single Image Novel View Synthesis via Test-Time Adaptation of Diffusion Models

Data Augmentation Through Random Style Replacement

DMPT: Decoupled Modality-aware Prompt Tuning for Multi-modal Object Re-identification

Learning Optimal Prompt Ensemble for Multi-source Visual Prompt Transfer

DMM: Building a Versatile Image Generation Model via Distillation-Based Model Merging

Pose and Facial Expression Transfer by using StyleGAN

Effective Dual-Region Augmentation for Reduced Reliance on Large Amounts of Labeled Data

Built with on top of