Ordinal regression aims to predict ordered classes. Existing methods mainly focus on label distribution shapes and feature distance relationships, while the directional characteristics in the representation space remain underexplored. In this paper, we propose deep orientational representation learning (ORL), aiming to ensure the trajectory of features sequentially connected by ordinal categories approximates a geodesic. We treat the output layer weights as ordinal prototypes and introduce two constraints, the co-directional constraint and the counter-directional constraint. They operate by constraining the angles between pairs of vectors. The former minimizes the angle between vectors with matching start and end categories, while the latter maximizes the angle between vectors whose start categories are the same but whose end categories are on opposite sides. The two constraints optimize the representation from different ordinal directions. ORL is extended to a multi-prototype setting (MORL) to mitigate misalignment between features and oriented prototypes caused by large intra-class variations. Theoretical analysis links ORL to distribution unimodality and distance orderliness, highlighting its advantages. The effectiveness of ORL (MORL) is demonstrated on various tasks including facial age estimation, historical image dating, and aesthetic quality assessment.

Image animation has seen significant progress, driven by the powerful generative capabilities of diffusion models. However, maintaining appearance consistency with static input images and mitigating abrupt motion transitions in generated animations remain persistent challenges. While text-to-video (T2V) generation has demonstrated impressive performance with diffusion transformer models, the image animation field still largely relies on U-Net-based diffusion models, which lag behind the latest T2V approaches. Moreover, the quadratic complexity of vanilla self-attention mechanisms in Transformers imposes heavy computational demands, making image animation particularly resource-intensive. To address these issues, we propose MiraMo, a framework designed to enhance efficiency, appearance consistency, and motion smoothness in image animation. Specifically, MiraMo introduces three key elements: (1) A foundational text-to-video architecture replacing vanilla self-attention with efficient linear attention to reduce computational overhead while preserving generation quality; (2) A novel motion residual learning paradigm that focuses on modeling motion dynamics rather than directly predicting frames, improving temporal consistency; and (3) A DCT-based noise refinement strategy during inference to suppress sudden motion artifacts, complemented by a dynamics control module to balance motion smoothness and expressiveness. Extensive experiments against state-of-the-art methods validate the superiority of MiraMo in generating consistent, smooth, and controllable animations with accelerated inference speed. Additionally, we demonstrate the versatility of MiraMo through applications in motion transfer and video editing tasks.

Diffusion models have achieved significant progress in the task of image animation due to their powerful generative capabilities. However, preserving appearance consistency to the static input image, and avoiding abrupt motion change in the generated animation, remains challenging. In this paper, we introduce Cinemo, a novel image animation approach that aims at achieving better appearance consistency and motion smoothness. The core of Cinemo is to focus on learning the distribution of motion residuals, rather than directly predicting frames as in existing diffusion models. During the inference, we further mitigate the sudden motion changes in the generated video by introducing a novel DCT-based noise refinement strategy. To counteract the over-smoothing of motion, we introduce a dynamics degree control design for better control of the magnitude of motion. Altogether, these strategies enable Cinemo to produce highly consistent, smooth, and motion-controllable results. Extensive experiments compared with several state-of-the-art methods demonstrate the effectiveness and superiority of our proposed approach. In the end, we also demonstrate how our model can be applied for motion transfer or video editing of any given video. The project page is available at https://maxin-cn.github.io/cinemo_project/.

Recently video generation has achieved substantial progress with realistic results. Nevertheless, existing AI-generated videos are usually very short clips ("shot-level'') depicting a single scene. To deliver a coherent long video ("story-level''), it is desirable to have creative transition and prediction effects across different clips. This paper presents a short-to-long video diffusion model, SEINE, that focuses on generative transition and prediction. The goal is to generate high-quality long videos with smooth and creative transitions between scenes and varying lengths of shot-level videos. Specifically, we propose a random-mask video diffusion model to automatically generate transitions based on textual descriptions. By providing the images of different scenes as inputs, combined with text-based control, our model generates transition videos that ensure coherence and visual quality. Furthermore, the model can be readily extended to various tasks such as image-to-video animation and autoregressive video prediction. To conduct a comprehensive evaluation of this new generative task, we propose three assessing criteria for smooth and creative transition: temporal consistency, semantic similarity, and video-text semantic alignment. Extensive experiments validate the effectiveness of our approach over existing methods for generative transition and prediction, enabling the creation of story-level long videos.

This paper introduces InternVid, a large-scale video-centric multimodal dataset that enables learning powerful and transferable video-text representations for multimodal understanding and generation. InternVid contains over 7 million videos lasting nearly 760K hours, yielding 234M video clips accompanied by detailed descriptions of total 4.1B words. Our core contribution is to develop a scalable approach to autonomously build a high-quality video-text dataset with large language models (LLM), thereby showcasing its efficacy in learning video-language representation at scale. Specifically, we utilize a multi-scale approach to generate video-related descriptions. Furthermore, we introduce ViCLIP, a video-text representation learning model based on ViT-L. Learned on InternVid via contrastive learning, this model demonstrates leading zero-shot action recognition and competitive video retrieval performance. Beyond basic video understanding tasks like recognition and retrieval, our dataset and model have broad applications. They are particularly beneficial for generating interleaved video-text data for learning a video-centric dialogue system, advancing video-to-text and text-to-video generation research. These proposed resources provide a tool for researchers and practitioners interested in multimodal video understanding and generation.

While most image inpainting methods perform well on small image defects, they still struggle to deliver satisfactory results on large holes due to insufficient image guidance. To address this challenge, this paper proposes an uncertainty-aware adaptive feedback network (U2AFN), which incorporates an adaptive feedback mechanism to refine inpainting regions progressively. U2AFN predicts both an uncertainty map and an inpainting result simultaneously. During each iteration, the adaptive integration feedback block utilizes inpainting pixels with low uncertainty to guide the subsequent learning iteration. This process leads to a gradual reduction in uncertainty and produces more reliable inpainting outcomes. Our approach is extensively evaluated and compared on multiple datasets, demonstrating its superior performance over existing methods. The code is available at: https://codeocean.com/capsule/1901983/tree.

Few-sample compression aims to compress a big redundant model into a small compact one with only few samples. If we fine-tune models with these limited few samples directly, models will be vulnerable to overfit and learn almost nothing. Hence, previous methods optimize the compressed model layer-by-layer and try to make every layer have the same outputs as the corresponding layer in the teacher model, which is cumbersome. In this paper, we propose a new framework named Mimicking then Replacing (MiR) for few-sample compression, which firstly urges the pruned model to output the same features as the teacher's in the penultimate layer, and then replaces teacher's layers before penultimate with a well-tuned compact one. Unlike previous layer-wise reconstruction methods, our MiR optimizes the entire network holistically, which is not only simple and effective, but also unsupervised and general. MiR outperforms previous methods with large margins. Codes is available at https://github.com/cjnjuwhy/MiR.

Substantial improvements have been achieved in the field of face recognition due to the successful application of deep neural networks. However, existing methods are sensitive to both the quality and quantity of the training data. Despite the availability of large-scale datasets, the long tail data distribution induces strong biases in model learning. In this paper, we present a Face Augmentation Generative Adversarial Network (FA-GAN) to reduce the influence of imbalanced deformation attribute distributions. We propose to decouple these attributes from the identity representation with a novel hierarchical disentanglement module. Moreover, Graph Convolutional Networks (GCNs) are applied to recover geometric information by exploring the interrelations among local regions to guarantee the preservation of identities in face data augmentation. Extensive experiments on face reconstruction, face manipulation, and face recognition demonstrate the effectiveness and generalization ability of the proposed method.

Near-infrared-visual (NIR-VIS) heterogeneous face recognition (HFR) aims to match NIR face images with the corresponding VIS ones. It is a challenging task due to the sensing gaps among different modalities. Occlusions in the input face images make the task extremely complex. To tackle these problems, we present a Saliency Search Network (SSN) to extract domain-invariant identity features. We propose to automatically search the efficient parts of face images in a modality-aware manner, and remove redundant information. Moreover, the searching process is guided by an information bottleneck network, which mitigates the overfitting problems caused by small datasets. Extensive experiments on both complete and partial NIR-VIS HFR on multiple datasets demonstrate the effectiveness and robustness of the proposed method to modality discrepancy and occlusions.

Photo-to-caricature translation aims to synthesize the caricature as a rendered image exaggerating the features through sketching, pencil strokes, or other artistic drawings. Style rendering and geometry deformation are the most important aspects in photo-to-caricature translation task. To take both into consideration, we propose an unsupervised contrastive photo-to-caricature translation architecture. Considering the intuitive artifacts in the existing methods, we propose a contrastive style loss for style rendering to enforce the similarity between the style of rendered photo and the caricature, and simultaneously enhance its discrepancy to the photos. To obtain an exaggerating deformation in an unpaired/unsupervised fashion, we propose a Distortion Prediction Module (DPM) to predict a set of displacements vectors for each input image while fixing some controlling points, followed by the thin plate spline interpolation for warping. The model is trained on unpaired photo and caricature while can offer bidirectional synthesizing via inputting either a photo or a caricature. Extensive experiments demonstrate that the proposed model is effective to generate hand-drawn like caricatures compared with existing competitors.

Most deep learning based image inpainting approaches adopt autoencoder or its variants to fill missing regions in images. Encoders are usually utilized to learn powerful representational spaces, which are important for dealing with sophisticated learning tasks. Specifically, in image inpainting tasks, masks with any shapes can appear anywhere in images (i.e., free-form masks) which form complex patterns. It is difficult for encoders to capture such powerful representations under this complex situation. To tackle this problem, we propose a self-supervised Siamese inference network to improve the robustness and generalization. It can encode contextual semantics from full resolution images and obtain more discriminative representations. we further propose a multi-scale decoder with a novel dual attention fusion module (DAF), which can combine both the restored and known regions in a smooth way. This multi-scale architecture is benefit for decoding discriminative representations learned by encoders into images layer by layer. In this way, unknown regions will be filled naturally from outside to inside. Qualitative and quantitative experiments on multiple datasets, including facial and natural datasets (i.e., Celeb-HQ, Pairs Street View, Places2 and ImageNet), demonstrate that our proposed method outperforms state-of-the-art methods in generating high-quality inpainting results.