Mengmeng Ge

Xiaomin Li, Xu Jia, Qinghe Wang et al.

Existing pretrained text-to-video (T2V) models have demonstrated impressive abilities in generating realistic videos with basic motion or camera movement. However, these models exhibit significant limitations when generating intricate, human-centric motions. Current efforts primarily focus on fine-tuning models on a small set of videos containing a specific motion. They often fail to effectively decouple motion and the appearance in the limited reference videos, thereby weakening the modeling capability of motion patterns. To this end, we propose MoTrans, a customized motion transfer method enabling video generation of similar motion in new context. Specifically, we introduce a multimodal large language model (MLLM)-based recaptioner to expand the initial prompt to focus more on appearance and an appearance injection module to adapt appearance prior from video frames to the motion modeling process. These complementary multimodal representations from recaptioned prompt and video frames promote the modeling of appearance and facilitate the decoupling of appearance and motion. In addition, we devise a motion-specific embedding for further enhancing the modeling of the specific motion. Experimental results demonstrate that our method effectively learns specific motion pattern from singular or multiple reference videos, performing favorably against existing methods in customized video generation.

Takashi Isobe, He Cui, Dong Zhou et al.

Text-to-Video (T2V) generation has attracted significant attention for its ability to synthesize realistic videos from textual descriptions. However, existing models struggle to balance computational efficiency and high visual quality, particularly on resource-limited devices, e.g.,iGPUs and mobile phones. Most prior work prioritizes visual fidelity while overlooking the need for smaller, more efficient models suitable for real-world deployment. To address this challenge, we propose a lightweight T2V framework, termed Hummingbird, which prunes existing models and enhances visual quality through visual feedback learning. Our approach reduces the size of the U-Net from 1.4 billion to 0.7 billion parameters, significantly improving efficiency while preserving high-quality video generation. Additionally, we introduce a novel data processing pipeline that leverages Large Language Models (LLMs) and Video Quality Assessment (VQA) models to enhance the quality of both text prompts and video data. To support user-driven training and style customization, we publicly release the full training code, including data processing and model training. Extensive experiments show that our method achieves a 31X speedup compared to state-of-the-art models such as VideoCrafter2, while also attaining the highest overall score on VBench. Moreover, our method supports the generation of videos with up to 26 frames, addressing the limitations of existing U-Net-based methods in long video generation. Notably, the entire training process requires only four GPUs, yet delivers performance competitive with existing leading methods. Hummingbird presents a practical and efficient solution for T2V generation, combining high performance, scalability, and flexibility for real-world applications.

Zirui Zheng, Takashi Isobe, Tong Shen et al.

While autoregressive (AR) models have demonstrated remarkable success in image generation, extending them to layout-conditioned generation remains challenging due to the sparse nature of layout conditions and the risk of feature entanglement. We present Structured Masking for AR-based Layout-to-Image (SMARLI), a novel framework for layoutto-image generation that effectively integrates spatial layout constraints into AR-based image generation. To equip AR model with layout control, a specially designed structured masking strategy is applied to attention computation to govern the interaction among the global prompt, layout, and image tokens. This design prevents mis-association between different regions and their descriptions while enabling sufficient injection of layout constraints into the generation process. To further enhance generation quality and layout accuracy, we incorporate Group Relative Policy Optimization (GRPO) based post-training scheme with specially designed layout reward functions for next-set-based AR models. Experimental results demonstrate that SMARLI is able to seamlessly integrate layout tokens with text and image tokens without compromising generation quality. It achieves superior layoutaware control while maintaining the structural simplicity and generation efficiency of AR models.

Sinan Tan, Mengmeng Ge, Di Guo et al.

In the Vision-and-Language Navigation task, the embodied agent follows linguistic instructions and navigates to a specific goal. It is important in many practical scenarios and has attracted extensive attention from both computer vision and robotics communities. However, most existing works only use RGB images but neglect the 3D semantic information of the scene. To this end, we develop a novel self-supervised training framework to encode the voxel-level 3D semantic reconstruction into a 3D semantic representation. Specifically, a region query task is designed as the pretext task, which predicts the presence or absence of objects of a particular class in a specific 3D region. Then, we construct an LSTM-based navigation model and train it with the proposed 3D semantic representations and BERT language features on vision-language pairs. Experiments show that the proposed approach achieves success rates of 68% and 66% on the validation unseen and test unseen splits of the R2R dataset respectively, which are superior to most of RGB-based methods utilizing vision-language transformers.

Sinan Tan, Mengmeng Ge, Di Guo et al.

In this paper, we propose a novel Knowledge-based Embodied Question Answering (K-EQA) task, in which the agent intelligently explores the environment to answer various questions with the knowledge. Different from explicitly specifying the target object in the question as existing EQA work, the agent can resort to external knowledge to understand more complicated question such as "Please tell me what are objects used to cut food in the room?", in which the agent must know the knowledge such as "knife is used for cutting food". To address this K-EQA problem, a novel framework based on neural program synthesis reasoning is proposed, where the joint reasoning of the external knowledge and 3D scene graph is performed to realize navigation and question answering. Especially, the 3D scene graph can provide the memory to store the visual information of visited scenes, which significantly improves the efficiency for the multi-turn question answering. Experimental results have demonstrated that the proposed framework is capable of answering more complicated and realistic questions in the embodied environment. The proposed method is also applicable to multi-agent scenarios.