7 papers
PISCES: Annotation-free Text-to-Video Post-Training via Optimal Transport-Aligned Rewards
Minh-Quan Le, Gaurav Mittal, Cheng Zhao +3
Text-to-video (T2V) generation aims to synthesize videos with high visual quality and temporal consistency that are semantically aligned with input text. Reward-based post-training…
Tracking-Guided 4D Generation: Foundation-Tracker Motion Priors for 3D Model Animation
Su Sun, Cheng Zhao, Himangi Mittal +4
Generating dynamic 4D objects from sparse inputs is difficult because it demands joint preservation of appearance and motion coherence across views and time while suppressing artif…
Scenes as Tokens: Multi-Scale Normal Distributions Transform Tokenizer for General 3D Vision-Language Understanding
Yutao Tang, Cheng Zhao, Gaurav Mittal +4
Recent advances in 3D vision-language models (VLMs) highlight a strong potential for 3D scene understanding and reasoning. However, effectively tokenizing 3D scenes into holistic s…
Sig-DEG for Distillation: Making Diffusion Models Faster and Lighter
Lei Jiang, Wen Ge, Niels Cariou-Kotlarek +6
Diffusion models have achieved state-of-the-art results in generative modelling but remain computationally intensive at inference time, often requiring thousands of discretization…
Hummingbird: High Fidelity Image Generation via Multimodal Context Alignment
Minh-Quan Le, Gaurav Mittal, Tianjian Meng +5
While diffusion models are powerful in generating high-quality, diverse synthetic data for object-centric tasks, existing methods struggle with scene-aware tasks such as Visual Que…
DeCafNet: Delegate and Conquer for Efficient Temporal Grounding in Long Videos
Zijia Lu, A S M Iftekhar, Gaurav Mittal +6
Long Video Temporal Grounding (LVTG) aims at identifying specific moments within lengthy videos based on user-provided text queries for effective content retrieval. The approach ta…