JPEG compression adopts the quantization of Discrete Cosine Transform (DCT) coefficients for effective bit-rate reduction, whilst the quantization could lead to a significant loss of important image details. Recovering compressed JPEG images in the frequency domain has attracted more and more attention recently, in addition to numerous restoration approaches developed in the pixel domain. However, the current DCT domain methods typically suffer from limited effectiveness in handling a wide range of compression quality factors, or fall short in recovering sparse quantized coefficients and the components across different colorspace. To address these challenges, we propose a DCT domain spatial-frequential Transformer, named as DCTransformer. Specifically, a dual-branch architecture is designed to capture both spatial and frequential correlations within the collocated DCT coefficients. Moreover, we incorporate the operation of quantization matrix embedding, which effectively allows our single model to handle a wide range of quality factors, and a luminance-chrominance alignment head that produces a unified feature map to align different-sized luminance and chrominance components. Our proposed DCTransformer outperforms the current state-of-the-art JPEG artifact removal techniques, as demonstrated by our extensive experiments.
Learned image compression methods have shown superior rate-distortion performance and remarkable potential compared to traditional compression methods. Most existing learned approaches use stacked convolution or window-based self-attention for transform coding, which aggregate spatial information in a fixed range. In this paper, we focus on extending spatial aggregation capability and propose a dynamic kernel-based transform coding. The proposed adaptive aggregation generates kernel offsets to capture valid information in the content-conditioned range to help transform. With the adaptive aggregation strategy and the sharing weights mechanism, our method can achieve promising transform capability with acceptable model complexity. Besides, according to the recent progress of entropy model, we define a generalized coarse-to-fine entropy model, considering the coarse global context, the channel-wise, and the spatial context. Based on it, we introduce dynamic kernel in hyper-prior to generate more expressive global context. Furthermore, we propose an asymmetric spatial-channel entropy model according to the investigation of the spatial characteristics of the grouped latents. The asymmetric entropy model aims to reduce statistical redundancy while maintaining coding efficiency. Experimental results demonstrate that our method achieves superior rate-distortion performance on three benchmarks compared to the state-of-the-art learning-based methods.
Pretrained visual-language models have demonstrated impressive zero-shot abilities in image captioning, when accompanied by hand-crafted prompts. Meanwhile, hand-crafted prompts utilize human prior knowledge to guide the model. However, due to the diversity between different domains, such hand-crafted prompt that provide invariant prior knowledge may result in mode collapse for some domains. Some researches attempted to incorporate expert knowledge and instruction datasets, but the results were costly and led to hallucinations. In this paper, we propose an unsupervised prompt learning method to improve Generalization of Image Captioning (GeneIC), which learns a domain-specific prompt vector for the target domain without requiring annotated data. GeneIC aligns visual and language modalities with a pre-trained Contrastive Language-Image Pre-Training (CLIP) model, thus optimizing the domain-specific prompt vector from two aspects: attribute and semantic consistency. Specifically, GeneIC first generates attribute-transferred images with differing attributes, while retaining semantic similarity with original images. Then, GeneIC uses CLIP to measure the similarity between the images and the generated sentences. By exploring the variable and invariant features in the original images and attribute-transferred images, attribute consistency constrains the attribute change direction of both images and sentences to learn domain-specific knowledge. The semantic consistency directly measures the similarity between the generated sentences and images to ensure the accuracy and comprehensiveness of the generated sentences. Consequently, GeneIC only optimizes the prompt vectors, which effectively retains the knowledge in the large model and introduces domain-specific knowledge.
Learned Image Compression (LIC) has recently become the trending technique for image transmission due to its notable performance. Despite its popularity, the robustness of LIC with respect to the quality of image reconstruction remains under-explored. In this paper, we introduce an imperceptible attack approach designed to effectively degrade the reconstruction quality of LIC, resulting in the reconstructed image being severely disrupted by noise where any object in the reconstructed images is virtually impossible. More specifically, we generate adversarial examples by introducing a Frobenius norm-based loss function to maximize the discrepancy between original images and reconstructed adversarial examples. Further, leveraging the insensitivity of high-frequency components to human vision, we introduce Imperceptibility Constraint (IC) to ensure that the perturbations remain inconspicuous. Experiments conducted on the Kodak dataset using various LIC models demonstrate effectiveness. In addition, we provide several findings and suggestions for designing future defenses.
Low-light image enhancement (LLIE) aims to improve the illuminance of images due to insufficient light exposure. Recently, various lightweight learning-based LLIE methods have been proposed to handle the challenges of unfavorable prevailing low contrast, low brightness, etc. In this paper, we have streamlined the architecture of the network to the utmost degree. By utilizing the effective structural re-parameterization technique, a single convolutional layer model (SCLM) is proposed that provides global low-light enhancement as the coarsely enhanced results. In addition, we introduce a local adaptation module that learns a set of shared parameters to accomplish local illumination correction to address the issue of varied exposure levels in different image regions. Experimental results demonstrate that the proposed method performs favorably against the state-of-the-art LLIE methods in both objective metrics and subjective visual effects. Additionally, our method has fewer parameters and lower inference complexity compared to other learning-based schemes.
Generating coherent and natural movement is the key challenge in video generation. This research proposes to condense video generation into a problem of motion generation, to improve the expressiveness of motion and make video generation more manageable. This can be achieved by breaking down the video generation process into latent motion generation and video reconstruction. We present a latent motion diffusion (LaMD) framework, which consists of a motion-decomposed video autoencoder and a diffusion-based motion generator, to implement this idea. Through careful design, the motion-decomposed video autoencoder can compress patterns in movement into a concise latent motion representation. Meanwhile, the diffusion-based motion generator is able to efficiently generate realistic motion on a continuous latent space under multi-modal conditions, at a cost that is similar to that of image diffusion models. Results show that LaMD generates high-quality videos with a wide range of motions, from stochastic dynamics to highly controllable movements. It achieves new state-of-the-art performance on benchmark datasets, including BAIR, Landscape and CATER-GENs, for Image-to-Video (I2V) and Text-Image-to-Video (TI2V) generation. The source code of LaMD will be made available soon.
In this paper, we consider the task of space-time video super-resolution (ST-VSR), namely, expanding a given source video to a higher frame rate and resolution simultaneously. However, most existing schemes either consider a fixed intermediate time and scale in the training stage or only accept a preset number of input frames (e.g., two adjacent frames) that fails to exploit long-range temporal information. To address these problems, we propose a continuous ST-VSR (C-STVSR) method that can convert the given video to any frame rate and spatial resolution. To achieve time-arbitrary interpolation, we propose a forward warping guided frame synthesis module and an optical-flow-guided context consistency loss to better approximate extreme motion and preserve similar structures among input and prediction frames. In addition, we design a memory-friendly cascading depth-to-space module to realize continuous spatial upsampling. Meanwhile, with the sophisticated reorganization of optical flow, the proposed method is memory friendly, making it possible to propagate information from long-range neighboring frames and achieve better reconstruction quality. Extensive experiments show that the proposed algorithm has good flexibility and achieves better performance on various datasets compared with the state-of-the-art methods in both objective evaluations and subjective visual effects.
Recently, user-oriented auto-encoders (UAEs) have been widely used in recommender systems to learn semantic representations of users based on their historical ratings. However, since latent item variables are not modeled in UAE, it is difficult to utilize the widely available item content information when ratings are sparse. In addition, whenever new items arrive, we need to wait for collecting rating data for these items and retrain the UAE from scratch, which is inefficient in practice. Aiming to address the above two problems simultaneously, we propose a mutually-regularized dual collaborative variational auto-encoder (MD-CVAE) for recommendation. First, by replacing randomly initialized last layer weights of the vanilla UAE with stacked latent item embeddings, MD-CVAE integrates two heterogeneous information sources, i.e., item content and user ratings, into the same principled variational framework where the weights of UAE are regularized by item content such that convergence to a non-optima due to data sparsity can be avoided. In addition, the regularization is mutual in that user ratings can also help the dual item content module learn more recommendation-oriented item content embeddings. Finally, we propose a symmetric inference strategy for MD-CVAE where the first layer weights of the UAE encoder are tied to the latent item embeddings of the UAE decoder. Through this strategy, no retraining is required to recommend newly introduced items. Empirical studies show the effectiveness of MD-CVAE in both normal and cold-start scenarios. Codes are available at https://github.com/yaochenzhu/MD-CVAE.
Image super-resolution is a common task on mobile and IoT devices, where one often needs to upscale and enhance low-resolution images and video frames. While numerous solutions have been proposed for this problem in the past, they are usually not compatible with low-power mobile NPUs having many computational and memory constraints. In this Mobile AI challenge, we address this problem and propose the participants to design an efficient quantized image super-resolution solution that can demonstrate a real-time performance on mobile NPUs. The participants were provided with the DIV2K dataset and trained INT8 models to do a high-quality 3X image upscaling. The runtime of all models was evaluated on the Synaptics VS680 Smart Home board with a dedicated edge NPU capable of accelerating quantized neural networks. All proposed solutions are fully compatible with the above NPU, demonstrating an up to 60 FPS rate when reconstructing Full HD resolution images. A detailed description of all models developed in the challenge is provided in this paper.
Next point-of-interest (POI) recommendation is a critical task in location-based social networks, yet remains challenging due to a high degree of variation and personalization exhibited in user movements. In this work, we explore the latent hierarchical structure composed of multi-granularity short-term structural patterns in user check-in sequences. We propose a Spatio-Temporal context AggRegated Hierarchical Transformer (STAR-HiT) for next POI recommendation, which employs stacked hierarchical encoders to recursively encode the spatio-temporal context and explicitly locate subsequences of different granularities. More specifically, in each encoder, the global attention layer captures the spatio-temporal context of the sequence, while the local attention layer performed within each subsequence enhances subsequence modeling using the local context. The sequence partition layer infers positions and lengths of subsequences from the global context adaptively, such that semantics in subsequences can be well preserved. Finally, the subsequence aggregation layer fuses representations within each subsequence to form the corresponding subsequence representation, thereby generating a new sequence of higher-level granularity. The stacking of encoders captures the latent hierarchical structure of the check-in sequence, which is used to predict the next visiting POI. Extensive experiments on three public datasets demonstrate that the proposed model achieves superior performance whilst providing explanations for recommendations. Codes are available at https://github.com/JennyXieJiayi/STAR-HiT.