Text-to-image users often provide concise and underspecified prompts, whereas generative models benefit from detailed textual conditions for reliable instruction following. Existing systems bridge this gap with Prompt Enhancers (PEs) that rewrite raw prompts at inference time, introducing additional latency and leaving...
Ming-Hung Lin, Cheng-Fei Cai, Lin Xu et al.· 0 citations
Recent advances in text-to-image (T2I) generation have enabled controllable image synthesis by incorporating conditions beyond text. However, most existing diffusion-based methods are limited to a single type of control condition (e.g., bounding boxes or keypoints), which restricts their flexibility. To address this li...
Pengcheng Wan, Liang Han, Lin Xu et al.· arXiv.org· 0 citations
This work proposes DreOPD, a Degraded-reference extrapolative OPD method for flow-matching models that bridges these two paradigms, and converts implicit reward extrapolation into closed-form velocity regression, enabling extrapolative post-training with the stability of OPD.
Ming-Hung Lin, Cheng-Fei Cai, Lin Xu et al.· 2 citations
A normal-guided depth propagation approach is introduced, which can extend depth information from high-confidence regions to constrain the depth in low-confidence areas and an abnormal depth edge-aware regularization is proposed to address depth discontinuities caused by the discreteness of Gaussians.
This paper proposes a novel yet simple partitioning method to efficiently and faithfully reconstruct large-scale scene surfaces and proposes a strategy to detect and repair missing regions in the initial point cloud caused by sparse viewpoints or insufficient textures, thereby further improving the geometric quality.
