In Learned Image Compression (LIC), a model is trained at encoding and decoding images sampled from a source domain, often outperforming traditional codecs on natural images; yet its performance may be far from optimal on images sampled from different domains. In this work, we tackle the problem of adapting a pre-trained model to multiple target domains by plugging into the decoder an adapter module for each of them, including the source one. Each adapter improves the decoder performance on a specific domain, without the model forgetting about the images seen at training time. A gate network computes the weights to optimally blend the contributions from the adapters when the bitstream is decoded. We experimentally validate our method over two state-of-the-art pre-trained models, observing improved rate-distortion efficiency on the target domains without penalties on the source domain. Furthermore, the gate's ability to find similarities with the learned target domains enables better encoding efficiency also for images outside them.
Feature pyramids have been widely adopted in convolutional neural networks (CNNs) and transformers for tasks like medical image segmentation and object detection. However, the currently existing models generally focus on the Encoder-side Transformer to extract features, from which decoder improvement can bring further potential with well-designed architecture. We propose CFPFormer, a novel decoder block that integrates feature pyramids and transformers. Specifically, by leveraging patch embedding, cross-layer feature concatenation, and Gaussian attention mechanisms, CFPFormer enhances feature extraction capabilities while promoting generalization across diverse tasks. Benefiting from Transformer structure and U-shaped Connections, our introduced model gains the ability to capture long-range dependencies and effectively up-sample feature maps. Our model achieves superior performance in detecting small objects compared to existing methods. We evaluate CFPFormer on medical image segmentation datasets and object detection benchmarks (VOC 2007, VOC2012, MS-COCO), demonstrating its effectiveness and versatility. On the ACDC Post-2017-MICCAI-Challenge online test set, our model reaches exceptionally impressive accuracy, and performed well compared with the original decoder setting in Synapse multi-organ segmentation dataset.
In-context learning (ICL) approaches typically leverage prompting to condition decoder-only language model generation on reference information. Just-in-time processing of a context is inefficient due to the quadratic cost of self-attention operations, and caching is desirable. However, caching transformer states can easily require almost as much space as the model parameters. When the right context isn't known in advance, caching ICL can be challenging. This work addresses these limitations by introducing models that, inspired by the encoder-decoder architecture, use cross-attention to condition generation on reference text without the prompt. More precisely, we leverage pre-trained decoder-only models and only train a small number of added layers. We use Question-Answering (QA) as a testbed to evaluate the ability of our models to perform conditional generation and observe that they outperform ICL, are comparable to fine-tuned prompted LLMs, and drastically reduce the space footprint relative to standard KV caching by two orders of magnitude.
To build large language models for Ukrainian we need to expand our corpora with large amounts of new algorithmic tasks expressed in natural language. Examples of task performance expressed in English are abundant, so with a high-quality translation system our community will be enabled to curate datasets faster. To aid this goal, we introduce a recipe to build a translation system using supervised finetuning of a large pretrained language model with a noisy parallel dataset of 3M pairs of Ukrainian and English sentences followed by a second phase of training using 17K examples selected by k-fold perplexity filtering on another dataset of higher quality. Our decoder-only model named Dragoman beats performance of previous state of the art encoder-decoder models on the FLORES devtest set.
Embedding-based retrieval serves as a dominant approach to candidate item matching for industrial recommender systems. With the success of generative AI, generative retrieval has recently emerged as a new retrieval paradigm for recommendation, which casts item retrieval as a generation problem. Its model consists of two stages: semantic tokenization and autoregressive generation. The first stage involves item tokenization that constructs discrete semantic tokens to index items, while the second stage autoregressively generates semantic tokens of candidate items. Therefore, semantic tokenization serves as a crucial preliminary step for training generative recommendation models. Existing research usually employs a vector quantizier with reconstruction loss (e.g., RQ-VAE) to obtain semantic tokens of items, but this method fails to capture the essential neighborhood relationships that are vital for effective item modeling in recommender systems. In this paper, we propose a contrastive quantization-based semantic tokenization approach, named CoST, which harnesses both item relationships and semantic information to learn semantic tokens. Our experimental results highlight the significant impact of semantic tokenization on generative recommendation performance, with CoST achieving up to a 43% improvement in Recall@5 and 44% improvement in NDCG@5 on the MIND dataset over previous baselines.
Intersection crossing represents one of the most dangerous sections of the road infrastructure and Connected Vehicles (CVs) can serve as a revolutionary solution to the problem. In this work, we present a novel framework that detects preemptively collisions at urban crossroads, exploiting the Multi-access Edge Computing (MEC) platform of 5G networks. At the MEC, an Intersection Manager (IM) collects information from both vehicles and the road infrastructure to create a holistic view of the area of interest. Based on the historical data collected, the IM leverages the capabilities of an encoder-decoder recurrent neural network to predict, with high accuracy, the future vehicles' trajectories. As, however, accuracy is not a sufficient measure of how much we can trust a model, trajectory predictions are additionally associated with a measure of uncertainty towards confident collision forecasting and avoidance. Hence, contrary to any other approach in the state of the art, an uncertainty-aware collision prediction framework is developed that is shown to detect well in advance (and with high reliability) if two vehicles are on a collision course. Subsequently, collision detection triggers a number of alarms that signal the colliding vehicles to brake. Under real-world settings, thanks to the preemptive capabilities of the proposed approach, all the simulated imminent dangers are averted.
In recent times, the research on Large Language Models (LLMs) has grown exponentially, predominantly focusing on models underpinned by the transformer architecture, as established by [1], and further developed through the decoder-only variations by [2]. Contemporary efforts in this field primarily aim to enhance model capabilities by scaling up both the architecture and data volumes utilized during training. However, the exploration into reduce these model sizes while preserving their efficacy remains scant. In this study, we introduce three modifications to the decoder-only transformer architecture, namely ParallelGPT (pgpt), LinearGPT (lgpt), and ConvGPT (cgpt). These variants demonstrate comparable performance to the conventional architecture in language generation, yet benefit from reduced model sizes and faster training processes. We open-source the model weights and the complete codebase for these implementation for further research.
Millimeter wave radar is gaining traction recently as a promising modality for enabling pervasive and privacy-preserving gesture recognition. However, the lack of rich and fine-grained radar datasets hinders progress in developing generalized deep learning models for gesture recognition across various user postures (e.g., standing, sitting), positions, and scenes. To remedy this, we resort to designing a software pipeline that exploits wealthy 2D videos to generate realistic radar data, but it needs to address the challenge of simulating diversified and fine-grained reflection properties of user gestures. To this end, we design G3R with three key components: (i) a gesture reflection point generator expands the arm's skeleton points to form human reflection points; (ii) a signal simulation model simulates the multipath reflection and attenuation of radar signals to output the human intensity map; (iii) an encoder-decoder model combines a sampling module and a fitting module to address the differences in number and distribution of points between generated and real-world radar data for generating realistic radar data. We implement and evaluate G3R using 2D videos from public data sources and self-collected real-world radar data, demonstrating its superiority over other state-of-the-art approaches for gesture recognition.
The wide deployment of Face Recognition (FR) systems poses privacy risks. One countermeasure is adversarial attack, deceiving unauthorized malicious FR, but it also disrupts regular identity verification of trusted authorizers, exacerbating the potential threat of identity impersonation. To address this, we propose the first double identity protection scheme based on traceable adversarial watermarking, termed DIP-Watermark. DIP-Watermark employs a one-time watermark embedding to deceive unauthorized FR models and allows authorizers to perform identity verification by extracting the watermark. Specifically, we propose an information-guided adversarial attack against FR models. The encoder embeds an identity-specific watermark into the deep feature space of the carrier, guiding recognizable features of the image to deviate from the source identity. We further adopt a collaborative meta-optimization strategy compatible with sub-tasks, which regularizes the joint optimization direction of the encoder and decoder. This strategy enhances the representation of universal carrier features, mitigating multi-objective optimization conflicts in watermarking. Experiments confirm that DIP-Watermark achieves significant attack success rates and traceability accuracy on state-of-the-art FR models, exhibiting remarkable robustness that outperforms the existing privacy protection methods using adversarial attacks and deep watermarking, or simple combinations of the two. Our work potentially opens up new insights into proactive protection for FR privacy.
Prior material creation methods had limitations in producing diverse results mainly because reconstruction-based methods relied on real-world measurements and generation-based methods were trained on relatively small material datasets. To address these challenges, we propose DreamPBR, a novel diffusion-based generative framework designed to create spatially-varying appearance properties guided by text and multi-modal controls, providing high controllability and diversity in material generation. Key to achieving diverse and high-quality PBR material generation lies in integrating the capabilities of recent large-scale vision-language models trained on billions of text-image pairs, along with material priors derived from hundreds of PBR material samples. We utilize a novel material Latent Diffusion Model (LDM) to establish the mapping between albedo maps and the corresponding latent space. The latent representation is then decoded into full SVBRDF parameter maps using a rendering-aware PBR decoder. Our method supports tileable generation through convolution with circular padding. Furthermore, we introduce a multi-modal guidance module, which includes pixel-aligned guidance, style image guidance, and 3D shape guidance, to enhance the control capabilities of the material LDM. We demonstrate the effectiveness of DreamPBR in material creation, showcasing its versatility and user-friendliness on a wide range of controllable generation and editing applications.
The paper introduces the UniMER dataset, marking the first study on Mathematical Expression Recognition (MER) targeting complex real-world scenarios. The UniMER dataset includes a large-scale training set, UniMER-1M, which offers unprecedented scale and diversity with one million training instances to train high-quality, robust models. Additionally, UniMER features a meticulously designed, diverse test set, UniMER-Test, which covers a variety of formula distributions found in real-world scenarios, providing a more comprehensive and fair evaluation. To better utilize the UniMER dataset, the paper proposes a Universal Mathematical Expression Recognition Network (UniMERNet), tailored to the characteristics of formula recognition. UniMERNet consists of a carefully designed encoder that incorporates detail-aware and local context features, and an optimized decoder for accelerated performance. Extensive experiments conducted using the UniMER-1M dataset and UniMERNet demonstrate that training on the large-scale UniMER-1M dataset can produce a more generalizable formula recognition model, significantly outperforming all previous datasets. Furthermore, the introduction of UniMERNet enhances the model's performance in formula recognition, achieving higher accuracy and speeds. All data, models, and code are available at https://github.com/opendatalab/UniMERNet.
Recently, implicit neural representations (INR) have made significant strides in various vision-related domains, providing a novel solution for Multispectral and Hyperspectral Image Fusion (MHIF) tasks. However, INR is prone to losing high-frequency information and is confined to the lack of global perceptual capabilities. To address these issues, this paper introduces a Fourier-enhanced Implicit Neural Fusion Network (FeINFN) specifically designed for MHIF task, targeting the following phenomena: The Fourier amplitudes of the HR-HSI latent code and LR-HSI are remarkably similar; however, their phases exhibit different patterns. In FeINFN, we innovatively propose a spatial and frequency implicit fusion function (Spa-Fre IFF), helping INR capture high-frequency information and expanding the receptive field. Besides, a new decoder employing a complex Gabor wavelet activation function, called Spatial-Frequency Interactive Decoder (SFID), is invented to enhance the interaction of INR features. Especially, we further theoretically prove that the Gabor wavelet activation possesses a time-frequency tightness property that favors learning the optimal bandwidths in the decoder. Experiments on two benchmark MHIF datasets verify the state-of-the-art (SOTA) performance of the proposed method, both visually and quantitatively. Also, ablation studies demonstrate the mentioned contributions. The code will be available on Anonymous GitHub (https://anonymous.4open.science/r/FeINFN-15C9/) after possible acceptance.
Salient object detection (SOD) aims at finding the most salient objects in images and outputs pixel-level binary masks. Transformer-based methods achieve promising performance due to their global semantic understanding, crucial for identifying salient objects. However, these models tend to be large and require numerous training parameters. To better harness the potential of transformers for SOD, we propose a novel parameter-efficient fine-tuning method aimed at reducing the number of training parameters while enhancing the salient object detection capability. Our model, termed EXternal Prompt features Enhanced adapteR Tuning (ExPert), features an encoder-decoder structure with adapters and injectors interspersed between the layers of a frozen transformer encoder. The adapter modules adapt the pretrained backbone to SOD while the injector modules incorporate external prompt features to enhance the awareness of salient objects. Comprehensive experiments demonstrate the superiority of our method. Surpassing former state-of-the-art (SOTA) models across five SOD datasets, ExPert achieves 0.215 mean absolute error (MAE) in the ECSSD dataset with 80.2M trained parameters, 21% better than SelfReformer and 47% better than EGNet.
Panoramic distortion poses a significant challenge in 360 depth estimation, particularly pronounced at the north and south poles. Existing methods either adopt a bi-projection fusion strategy to remove distortions or model long-range dependencies to capture global structures, which can result in either unclear structure or insufficient local perception. In this paper, we propose a spherical geometry transformer, named SGFormer, to address the above issues, with an innovative step to integrate spherical geometric priors into vision transformers. To this end, we retarget the transformer decoder to a spherical prior decoder (termed SPDecoder), which endeavors to uphold the integrity of spherical structures during decoding. Concretely, we leverage bipolar re-projection, circular rotation, and curve local embedding to preserve the spherical characteristics of equidistortion, continuity, and surface distance, respectively. Furthermore, we present a query-based global conditional position embedding to compensate for spatial structure at varying resolutions. It not only boosts the global perception of spatial position but also sharpens the depth structure across different patches. Finally, we conduct extensive experiments on popular benchmarks, demonstrating our superiority over state-of-the-art solutions.
Precisely grasping and reconstructing articulated objects is key to enabling general robotic manipulation. In this paper, we propose CenterArt, a novel approach for simultaneous 3D shape reconstruction and 6-DoF grasp estimation of articulated objects. CenterArt takes RGB-D images of the scene as input and first predicts the shape and joint codes through an encoder. The decoder then leverages these codes to reconstruct 3D shapes and estimate 6-DoF grasp poses of the objects. We further develop a mechanism for generating a dataset of 6-DoF grasp ground truth poses for articulated objects. CenterArt is trained on realistic scenes containing multiple articulated objects with randomized designs, textures, lighting conditions, and realistic depths. We perform extensive experiments demonstrating that CenterArt outperforms existing methods in accuracy and robustness.
This paper aims to conduct a comprehensive theoretical analysis of current diffusion models. We introduce a novel generative learning methodology utilizing the Schr{\"o}dinger bridge diffusion model in latent space as the framework for theoretical exploration in this domain. Our approach commences with the pre-training of an encoder-decoder architecture using data originating from a distribution that may diverge from the target distribution, thus facilitating the accommodation of a large sample size through the utilization of pre-existing large-scale models. Subsequently, we develop a diffusion model within the latent space utilizing the Schr{\"o}dinger bridge framework. Our theoretical analysis encompasses the establishment of end-to-end error analysis for learning distributions via the latent Schr{\"o}dinger bridge diffusion model. Specifically, we control the second-order Wasserstein distance between the generated distribution and the target distribution. Furthermore, our obtained convergence rates effectively mitigate the curse of dimensionality, offering robust theoretical support for prevailing diffusion models.
Dimension reduction techniques usually lose information in the sense that reconstructed data are not identical to the original data. However, we argue that it is possible to have reconstructed data identically distributed as the original data, irrespective of the retained dimension or the specific mapping. This can be achieved by learning a distributional model that matches the conditional distribution of data given its low-dimensional latent variables. Motivated by this, we propose Distributional Principal Autoencoder (DPA) that consists of an encoder that maps high-dimensional data to low-dimensional latent variables and a decoder that maps the latent variables back to the data space. For reducing the dimension, the DPA encoder aims to minimise the unexplained variability of the data with an adaptive choice of the latent dimension. For reconstructing data, the DPA decoder aims to match the conditional distribution of all data that are mapped to a certain latent value, thus ensuring that the reconstructed data retains the original data distribution. Our numerical results on climate data, single-cell data, and image benchmarks demonstrate the practical feasibility and success of the approach in reconstructing the original distribution of the data. DPA embeddings are shown to preserve meaningful structures of data such as the seasonal cycle for precipitations and cell types for gene expression.
This paper investigates the challenging problem of learned image compression (LIC) with extreme low bitrates. Previous LIC methods based on transmitting quantized continuous features often yield blurry and noisy reconstruction due to the severe quantization loss. While previous LIC methods based on learned codebooks that discretize visual space usually give poor-fidelity reconstruction due to the insufficient representation power of limited codewords in capturing faithful details. We propose a novel dual-stream framework, HyrbidFlow, which combines the continuous-feature-based and codebook-based streams to achieve both high perceptual quality and high fidelity under extreme low bitrates. The codebook-based stream benefits from the high-quality learned codebook priors to provide high quality and clarity in reconstructed images. The continuous feature stream targets at maintaining fidelity details. To achieve the ultra low bitrate, a masked token-based transformer is further proposed, where we only transmit a masked portion of codeword indices and recover the missing indices through token generation guided by information from the continuous feature stream. We also develop a bridging correction network to merge the two streams in pixel decoding for final image reconstruction, where the continuous stream features rectify biases of the codebook-based pixel decoder to impose reconstructed fidelity details. Experimental results demonstrate superior performance across several datasets under extremely low bitrates, compared with existing single-stream codebook-based or continuous-feature-based LIC methods.
Extracting table contents from documents such as scientific papers and financial reports and converting them into a format that can be processed by large language models is an important task in knowledge information processing. End-to-end approaches, which recognize not only table structure but also cell contents, achieved performance comparable to state-of-the-art models using external character recognition systems, and have potential for further improvements. In addition, these models can now recognize long tables with hundreds of cells by introducing local attention. However, the models recognize table structure in one direction from the header to the footer, and cell content recognition is performed independently for each cell, so there is no opportunity to retrieve useful information from the neighbor cells. In this paper, we propose a multi-cell content decoder and bidirectional mutual learning mechanism to improve the end-to-end approach. The effectiveness is demonstrated on two large datasets, and the experimental results show comparable performance to state-of-the-art models, even for long tables with large numbers of cells.
Semantic segmentation plays a crucial role in enabling comprehensive scene understanding for robotic systems. However, generating annotations is challenging, requiring labels for every pixel in an image. In scenarios like autonomous driving, there's a need to progressively incorporate new classes as the operating environment of the deployed agent becomes more complex. For enhanced annotation efficiency, ideally, only pixels belonging to new classes would be annotated. This approach is known as Continual Semantic Segmentation (CSS). Besides the common problem of classical catastrophic forgetting in the continual learning setting, CSS suffers from the inherent ambiguity of the background, a phenomenon we refer to as the "background shift'', since pixels labeled as background could correspond to future classes (forward background shift) or previous classes (backward background shift). As a result, continual learning approaches tend to fail. This paper proposes a Backward Background Shift Detector (BACS) to detect previously observed classes based on their distance in the latent space from the foreground centroids of previous steps. Moreover, we propose a modified version of the cross-entropy loss function, incorporating the BACS detector to down-weight background pixels associated with formerly observed classes. To combat catastrophic forgetting, we employ masked feature distillation alongside dark experience replay. Additionally, our approach includes a transformer decoder capable of adjusting to new classes without necessitating an additional classification head. We validate BACS's superior performance over existing state-of-the-art methods on standard CSS benchmarks.