Human-centric motion control in video generation remains a critical challenge, particularly when jointly controlling camera movements and human poses in scenarios like the iconic Grammy Glambot moment. While recent video diffusion models have made significant progress, existing approaches struggle with limited motion representations and inadequate integration of camera and human motion controls. In this work, we present TokenMotion, the first DiT-based video diffusion framework that enables fine-grained control over camera motion, human motion, and their joint interaction. We represent camera trajectories and human poses as spatio-temporal tokens to enable local control granularity. Our approach introduces a unified modeling framework utilizing a decouple-and-fuse strategy, bridged by a human-aware dynamic mask that effectively handles the spatially-and-temporally varying nature of combined motion signals. Through extensive experiments, we demonstrate TokenMotion’s effectiveness across both text-to-video and image-to-video paradigms, consistently outperforming current state-of-the-art methods in human-centric motion control tasks. Our work represents a significant advancement in controllable video generation, with particular relevance for creative production applications.

Video summarization aims to distill the most important information from a source video into either an abridged video clip or a textual narrative. Existing methods often treat the generation of video and text summaries as independent tasks, thus neglecting the semantic correlation between visual and textual summarization. In other words, these methods only study a single modality as output without considering coherent video and text as outputs. In this work, we first introduce a novel task: cross-modal video summarization. This task seeks to transfer a long video into a condensed video clip and a semantically aligned textual summary, collectively referred to as a cross-modal summary. We then establish VideoXum (X refers to different modalities), a new large-scale human-annotated video benchmark for cross-modal video summarization. VideoXum is reannotated based on ActivityNet Captions with diverse open-domain videos. In the current version, VideoXum provides 14K long videos, with a total of 140K pairs of aligned video and text summaries. Compared to existing datasets, VideoXum offers superior scalability while preserving a comparable level of annotation quality. To validate the dataset's quality, we provide a comprehensive analysis of VideoXum, comparing it with existing datasets. Further, we perform an extensive empirical evaluation of several state-of-the-art methods on this dataset. Our findings highlight the impressive generalization capability of the vision-language encoder-decoder framework yields on VideoXum. Particularly, we propose VTSUM-BLIP, an end-to-end framework, serving as a strong baseline for this novel benchmark. Moreover, we adapt CLIPScore for VideoXum to measure the semantic consistency of cross-modal summaries effectively. The project page ishttps://videoxum.github.io/.

The exploration of mutual-benefit cross-domains has shown great potential toward accurate self-supervised depth estimation. In this work, we revisit feature fusion between depth and semantic information and propose an efficient local adaptive attention method for geometric aware representation enhancement. Instead of building global connections or deforming attention across the feature space without restraint, we bound the spatial interaction within a learnable region of interest. In particular, we leverage geometric cues from semantic information to learn local adaptive bounding boxes to guide unsupervised feature aggregation. The local areas preclude most irrelevant reference points from attention space, yielding more selective feature learning and faster convergence. We naturally extend the paradigm into a multi-head and hierarchic way to enable the information distillation in different semantic levels and improve the feature discriminative ability for fine-grained depth estimation. Extensive experiments on the KITTI dataset show that our proposed method establishes a new state-of-the-art in self-supervised monocular depth estimation task, demonstrating the effectiveness of our approach over former Transformer variants.

The popular VQ-VAE models reconstruct images through learning a discrete codebook but suffer from a significant issue in the rapid quality degradation of image reconstruction as the compression rate rises. One major reason is that a higher compression rate induces more loss of visual signals on the higher frequency spectrum which reflect the details on pixel space. In this paper, a Frequency Complement Module (FCM) architecture is proposed to capture the missing frequency information for enhancing reconstruction quality. The FCM can be easily incorporated into the VQ-VAE structure, and we refer to the new model as Frequancy Augmented VAE (FAVAE). In addition, a Dynamic Spectrum Loss (DSL) is introduced to guide the FCMs to balance between various frequencies dynamically for optimal reconstruction. FA-VAE is further extended to the text-to-image synthesis task, and a Crossattention Autoregressive Transformer (CAT) is proposed to obtain more precise semantic attributes in texts. Extensive reconstruction experiments with different compression rates are conducted on several benchmark datasets, and the results demonstrate that the proposed FA-VAE is able to restore more faithfully the details compared to SOTA methods. CAT also shows improved generation quality with better image-text semantic alignment.

Semi-supervised dense prediction tasks, such as semantic segmentation, can be greatly improved through the use of contrastive learning. However, this approach presents two key challenges: selecting informative negative samples from a highly redundant pool and implementing effective data augmentation. To address these challenges, we present an adversarial contrastive learning method specifically for semi-supervised semantic segmentation. Direct learning of adversarial negatives is adopted to retain discriminative information from the past, leading to higher learning efficiency. Our approach also leverages an advanced data augmentation strategy called AdverseMix, which combines information from under-performing classes to generate more diverse and challenging samples. Additionally, we use auxiliary labels and classifiers to prevent over-adversarial negatives from affecting the learning process. Our experiments on the Pascal VOC and Cityscapes datasets demonstrate that our method outperforms the state-of-the-art by a significant margin, even when using a small fraction of labeled data.

Video-text retrieval is an essential task in cross-modal information retrieval, i.e., retrieving relevant videos from a large and unlabelled dataset given textual queries. Existing methods that simply pool the image features (e.g., based on the CLIP encoder [14]) from frames to build the video descriptor often result in sub-optimal video-text search accuracy since the information among different modalities is not fully exchanged and aligned. In this paper, we proposed a novel dual-encoder model to address the challenging video-text retrieval problem, which uses a highly efficient cross-attention module to facilitate the information exchange between multiple modalities (i.e., video and text). The proposed VideoCLIP is evaluated on two benchmark video-text datasets, MSRVTT and DiDeMo, and the results show that our model can outperform existing state-of-the-art methods while the retrieval speed is much faster than the traditional query-agnostic search model.

With the ever-increasing multimedia data on the Web, cross-modal video-text retrieval has received a lot of attention in recent years. Deep cross-modal hashing approaches utilize the Hamming space for achieving fast retrieval. However, most existing algorithms have difficulties in seeking or constructing a well-defined joint semantic space. In this paper, an unsupervised deep cross-modal video-text hashing approach (CLIP4Hashing) is proposed, which mitigates the difficulties in bridging between different modalities in the Hamming space through building a single hashing net by employing the pre-trained CLIP model. The approach is enhanced by two novel techniques, the dynamic weighting strategy and the design of the min-max hashing layer, which are found to be the main sources of the performance gain. Compared with conventional deep cross-modal hashing algorithms, CLIP4Hashing does not require data-specific hyper-parameters. With evaluation using three challenging video-text benchmark datasets, we demonstrate that CLIP4Hashing is able to significantly outperform existing state-of-the-art hashing algorithms. Additionally, with larger bit sizes (e.g., 2048 bits), CLIP4Hashing can even deliver competitive performance compared with the results based on non-hashing features.