Advancing Fine-Grained Visual Understanding with Multi-Scale Alignment in Multi-Modal Models

Multi-modal large language models (MLLMs) have achieved remarkable success in fine-grained visual understanding across a range of tasks. However, they often encounter significant challenges due to inadequate alignment for fine-grained knowledge, which restricts their ability to accurately capture local details and attain a comprehensive global perception. While recent advancements have focused on aligning object expressions with grounding information, they typically lack explicit integration of object images, which contain affluent information beyond mere texts or coordinates. To bridge this gap, we introduce a novel fine-grained visual knowledge alignment method that effectively aligns and integrates multi-scale knowledge of objects, including texts, coordinates, and images. This innovative method is underpinned by our multi-scale fine-grained enhancement data synthesis pipeline, which provides over 300K essential training data to enhance alignment and improve overall performance. Furthermore, we present TinyGroundingGPT, a series of compact models optimized for high-level alignments. With a scale of approximately 3B parameters, TinyGroundingGPT achieves outstanding results in grounding tasks while delivering performance comparable to larger MLLMs in complex visual scenarios.

SuperCodec: A Neural Speech Codec with Selective Back-Projection Network

Neural speech coding is a rapidly developing topic, where state-of-the-art approaches now exhibit superior compression performance than conventional methods. Despite significant progress, existing methods still have limitations in preserving and reconstructing fine details for optimal reconstruction, especially at low bitrates. In this study, we introduce SuperCodec, a neural speech codec that achieves state-of-the-art performance at low bitrates. It employs a novel back projection method with selective feature fusion for augmented representation. Specifically, we propose to use Selective Up-sampling Back Projection (SUBP) and Selective Down-sampling Back Projection (SDBP) modules to replace the standard up- and down-sampling layers at the encoder and decoder, respectively. Experimental results show that our method outperforms the existing neural speech codecs operating at various bitrates. Specifically, our proposed method can achieve higher quality reconstructed speech at 1 kbps than Lyra V2 at 3.2 kbps and Encodec at 6 kbps.

SimuSOE: A Simulated Snoring Dataset for Obstructive Sleep Apnea-Hypopnea Syndrome Evaluation during Wakefulness

Obstructive Sleep Apnea-Hypopnea Syndrome (OSAHS) is a prevalent chronic breathing disorder caused by upper airway obstruction. Previous studies advanced OSAHS evaluation through machine learning-based systems trained on sleep snoring or speech signal datasets. However, constructing datasets for training a precise and rapid OSAHS evaluation system poses a challenge, since 1) it is time-consuming to collect sleep snores and 2) the speech signal is limited in reflecting upper airway obstruction. In this paper, we propose a new snoring dataset for OSAHS evaluation, named SimuSOE, in which a novel and time-effective snoring collection method is introduced for tackling the above problems. In particular, we adopt simulated snoring which is a type of snore intentionally emitted by patients to replace natural snoring. Experimental results indicate that the simulated snoring signal during wakefulness can serve as an effective feature in OSAHS preliminary screening.

Aligning in a Compact Space: Contrastive Knowledge Distillation between Heterogeneous Architectures

Knowledge distillation is commonly employed to compress neural networks, reducing the inference costs and memory footprint. In the scenario of homogenous architecture, feature-based methods have been widely validated for their effectiveness. However, in scenarios where the teacher and student models are of heterogeneous architectures, the inherent differences in feature representation significantly degrade the performance of these methods. Recent studies have highlighted that low-frequency components constitute the majority of image features. Motivated by this, we propose a Low-Frequency Components-based Contrastive Knowledge Distillation (LFCC) framework that significantly enhances the performance of feature-based distillation between heterogeneous architectures. Specifically, we designe a set of multi-scale low-pass filters to extract the low-frequency components of intermediate features from both the teacher and student models, aligning them in a compact space to overcome architectural disparities. Moreover, leveraging the intrinsic pairing characteristic of the teacher-student framework, we design an innovative sample-level contrastive learning framework that adeptly restructures the constraints of within-sample feature similarity and between-sample feature divergence into a contrastive learning task. This strategy enables the student model to capitalize on intra-sample feature congruence while simultaneously enhancing the discrimination of features among disparate samples. Consequently, our LFCC framework accurately captures the commonalities in feature representation across heterogeneous architectures. Extensive evaluations and empirical analyses across three architectures (CNNs, Transformers, and MLPs) demonstrate that LFCC achieves superior performance on the challenging benchmarks of ImageNet-1K and CIFAR-100. All codes will be publicly available.

SAMDA: Leveraging SAM on Few-Shot Domain Adaptation for Electronic Microscopy Segmentation

It has been shown that traditional deep learning methods for electronic microscopy segmentation usually suffer from low transferability when samples and annotations are limited, while large-scale vision foundation models are more robust when transferring between different domains but facing sub-optimal improvement under fine-tuning. In this work, we present a new few-shot domain adaptation framework SAMDA, which combines the Segment Anything Model(SAM) with nnUNet in the embedding space to achieve high transferability and accuracy. Specifically, we choose the Unet-based network as the "expert" component to learn segmentation features efficiently and design a SAM-based adaptation module as the "generic" component for domain transfer. By amalgamating the "generic" and "expert" components, we mitigate the modality imbalance in the complex pre-training knowledge inherent to large-scale Vision Foundation models and the challenge of transferability inherent to traditional neural networks. The effectiveness of our model is evaluated on two electron microscopic image datasets with different modalities for mitochondria segmentation, which improves the dice coefficient on the target domain by 6.7%. Also, the SAM-based adaptor performs significantly better with only a single annotated image than the 10-shot domain adaptation on nnUNet. We further verify our model on four MRI datasets from different sources to prove its generalization ability.

Exploring General Intelligence via Gated Graph Transformer in Functional Connectivity Studies

Functional connectivity (FC) as derived from fMRI has emerged as a pivotal tool in elucidating the intricacies of various psychiatric disorders and delineating the neural pathways that underpin cognitive and behavioral dynamics inherent to the human brain. While Graph Neural Networks (GNNs) offer a structured approach to represent neuroimaging data, they are limited by their need for a predefined graph structure to depict associations between brain regions, a detail not solely provided by FCs. To bridge this gap, we introduce the Gated Graph Transformer (GGT) framework, designed to predict cognitive metrics based on FCs. Empirical validation on the Philadelphia Neurodevelopmental Cohort (PNC) underscores the superior predictive prowess of our model, further accentuating its potential in identifying pivotal neural connectivities that correlate with human cognitive processes.

Robust Multidimentional Chinese Remainder Theorem for Integer Vector Reconstruction

The problem of robustly reconstructing an integer vector from its erroneous remainders appears in many applications in the field of multidimensional (MD) signal processing. To address this problem, a robust MD Chinese remainder theorem (CRT) was recently proposed for a special class of moduli, where the remaining integer matrices left-divided by a greatest common left divisor (gcld) of all the moduli are pairwise commutative and coprime. The strict constraint on the moduli limits the usefulness of the robust MD-CRT in practice. In this paper, we investigate the robust MD-CRT for a general set of moduli. We first introduce a necessary and sufficient condition on the difference between paired remainder errors, followed by a simple sufficient condition on the remainder error bound, for the robust MD-CRT for general moduli, where the conditions are associated with (the minimum distances of) these lattices generated by gcld's of paired moduli, and a closed-form reconstruction algorithm is presented. We then generalize the above results of the robust MD-CRT from integer vectors/matrices to real ones. Finally, we validate the robust MD-CRT for general moduli by employing numerical simulations, and apply it to MD sinusoidal frequency estimation based on multiple sub-Nyquist samplers.

CDR-Adapter: Learning Adapters to Dig Out More Transferring Ability for Cross-Domain Recommendation Models

Data sparsity and cold-start problems are persistent challenges in recommendation systems. Cross-domain recommendation (CDR) is a promising solution that utilizes knowledge from the source domain to improve the recommendation performance in the target domain. Previous CDR approaches have mainly followed the Embedding and Mapping (EMCDR) framework, which involves learning a mapping function to facilitate knowledge transfer. However, these approaches necessitate re-engineering and re-training the network structure to incorporate transferrable knowledge, which can be computationally expensive and may result in catastrophic forgetting of the original knowledge. In this paper, we present a scalable and efficient paradigm to address data sparsity and cold-start issues in CDR, named CDR-Adapter, by decoupling the original recommendation model from the mapping function, without requiring re-engineering the network structure. Specifically, CDR-Adapter is a novel plug-and-play module that employs adapter modules to align feature representations, allowing for flexible knowledge transfer across different domains and efficient fine-tuning with minimal training costs. We conducted extensive experiments on the benchmark dataset, which demonstrated the effectiveness of our approach over several state-of-the-art CDR approaches.

MASTERKEY: Practical Backdoor Attack Against Speaker Verification Systems

Speaker Verification (SV) is widely deployed in mobile systems to authenticate legitimate users by using their voice traits. In this work, we propose a backdoor attack MASTERKEY, to compromise the SV models. Different from previous attacks, we focus on a real-world practical setting where the attacker possesses no knowledge of the intended victim. To design MASTERKEY, we investigate the limitation of existing poisoning attacks against unseen targets. Then, we optimize a universal backdoor that is capable of attacking arbitrary targets. Next, we embed the speaker's characteristics and semantics information into the backdoor, making it imperceptible. Finally, we estimate the channel distortion and integrate it into the backdoor. We validate our attack on 6 popular SV models. Specifically, we poison a total of 53 models and use our trigger to attack 16,430 enrolled speakers, composed of 310 target speakers enrolled in 53 poisoned models. Our attack achieves 100% attack success rate with a 15% poison rate. By decreasing the poison rate to 3%, the attack success rate remains around 50%. We validate our attack in 3 real-world scenarios and successfully demonstrate the attack through both over-the-air and over-the-telephony-line scenarios.

PhantomSound: Black-Box, Query-Efficient Audio Adversarial Attack via Split-Second Phoneme Injection

In this paper, we propose PhantomSound, a query-efficient black-box attack toward voice assistants. Existing black-box adversarial attacks on voice assistants either apply substitution models or leverage the intermediate model output to estimate the gradients for crafting adversarial audio samples. However, these attack approaches require a significant amount of queries with a lengthy training stage. PhantomSound leverages the decision-based attack to produce effective adversarial audios, and reduces the number of queries by optimizing the gradient estimation. In the experiments, we perform our attack against 4 different speech-to-text APIs under 3 real-world scenarios to demonstrate the real-time attack impact. The results show that PhantomSound is practical and robust in attacking 5 popular commercial voice controllable devices over the air, and is able to bypass 3 liveness detection mechanisms with >95% success rate. The benchmark result shows that PhantomSound can generate adversarial examples and launch the attack in a few minutes. We significantly enhance the query efficiency and reduce the cost of a successful untargeted and targeted adversarial attack by 93.1% and 65.5% compared with the state-of-the-art black-box attacks, using merely ~300 queries (~5 minutes) and ~1,500 queries (~25 minutes), respectively.