IJCAI 2024 Papers

3D semantic scene completion (SSC) is an ill-posed perception task that requires inferring a dense 3D scene from limited observations. Previous camera-based methods struggle to predict accurate semantic scenes due to inherent geometric ambiguity and incomplete observations. In this paper, we resort to stereo matching technique and bird’s-eye-view (BEV) representation learning to address such issues in SSC. Complementary to each other, stereo matching mitigates geometric ambiguity with epipolar constraint while BEV representation enhances the hallucination ability for invisible regions with global semantic context. However, due to the inherent representation gap between stereo geometry and BEV features, it is non-trivial to bridge them for dense prediction task of SSC. Therefore, we further develop a unified occupancy-based framework dubbed BRGScene, which effectively bridges these two representations with dense 3D volumes for reliable semantic scene completion. Specifically, we design a novel Mutual Interactive Ensemble (MIE) block for pixel-level reliable aggregation of stereo geometry and BEV features. Within the MIE block, a Bi-directional Reliable Interaction (BRI) module, enhanced with confidence re-weighting, is employed to encourage fine-grained interaction through mutual guidance. Besides, a Dual Volume Ensemble (DVE) module is introduced to facilitate complementary aggregation through channel-wise recalibration and multi-group voting. Our method outperforms all published camera-based methods on SemanticKITTI for semantic scene completion. Our code is available on https://github.com/Arlo0o/StereoScene.

We focus on the domain adaptation problem for 3D semantic segmentation, addressing the challenge of data variability in point clouds collected by different LiDARs. Existing benchmarks often mix different types of datasets, which blurs and complicates segmentation evaluations. Here, we introduce a Multi-LiDARs Domain Adaptation Segmentation (MLDAS) dataset, which contains point-wise semantic annotated point clouds captured simultaneously by a 128-beam LiDAR, a 64-beam LiDAR, a 32-beam LiDAR. We select 31,875 scans from 2 representative scenarios: campus and urban street. Furthermore, we evaluate the current 3D segmentation unsupervised domain adaptation methods on the proposed dataset and propose Hierarchical Segmentation Network with Spatial Consistency (HSSC) as a novel knowledge transfer method to mitigate the domain gap significantly using spatial-temporal consistency constraints. Extensive experiments show that HSSC greatly improves the state-of-the-art cross-domain semantic segmentation methods. Our project is available at https://sychen320.github.io/projects/MLDAS.

The remarkable prowess of diffusion models in image generation has spurred efforts to extend their application beyond generative tasks. However, a persistent challenge exists in lacking a unified approach to apply diffusion models to visual perception tasks with diverse semantic granularity requirements. Our purpose is to establish a unified visual perception framework, capitalizing on the potential synergies between generative and discriminative models. In this paper, we propose Vermouth, a simple yet effective framework comprising a pre-trained Stable Diffusion (SD) model containing rich generative priors, a unified head (U-head) capable of integrating hierarchical representations, and an Adapted-Expert providing discriminative priors. Comprehensive investigations unveil potential characteristics of Vermouth, such as varying granularity of perception concealed in latent variables at distinct time steps and various U-net stages. We emphasize that there is no necessity for incorporating a heavyweight or intricate decoder to transform diffusion models into potent representation learners. Extensive comparative evaluations against tailored discriminative models showcase the efficacy of our approach on zero-shot sketch-based image retrieval (ZS-SBIR), few-shot classification, and open-vocabulary (OV) semantic segmentation tasks. The promising results demonstrate the potential of diffusion models as formidable learners, establishing their significance in furnishing informative and robust visual representations.

Cued Speech (CS) is an advanced visual phonetic encoding system that integrates lip reading with hand codings, enabling people with hearing impairments to communicate efficiently. CS video generation aims to produce specific lip and gesture movements of CS from audio or text inputs. The main challenge is that given limited CS data, we strive to simultaneously generate fine-grained hand and finger movements, as well as lip movements, meanwhile the two kinds of movements need to be asynchronously aligned. Existing CS generation methods are fragile and prone to poor performance due to template-based statistical models and careful hand-crafted pre-processing to fit the models. Therefore, we propose a novel Gloss-prompted Diffusion-based CS Gesture generation framework (called GlossDiff). Specifically, to integrate additional linguistic rules knowledge into the model. we first introduce a bridging instruction called Gloss, which is an automatically generated descriptive text to establish a direct and more delicate semantic connection between spoken language and CS gestures. Moreover, we first suggest rhythm is an important paralinguistic feature for CS to improve the communication efficacy. Therefore, we propose a novel Audio-driven Rhythmic Module (ARM) to learn rhythm that matches audio speech. Moreover, in this work, we design, record, and publish the first Chinese CS dataset with four CS cuers. Extensive experiments demonstrate that our method quantitatively and qualitatively outperforms current state-of-the-art (SOTA) methods. We will release the code and data at glossdiff.github.io/.

Internet of Things (IoT) devices possess valuable yet private multimodal data, calling for a decentralized machine learning scheme. Though several multimodal federated learning (MFL) methods have been proposed, most of them merely overlook the system heterogeneity across IoT devices, resulting in the inadaptability to real world applications. Aiming at this, we conduct theoretical analysis and exploration experiments on straggler impacts and uncover the fact that stragglers caused by system heterogeneity are fatal to MFL, resulting in catastrophic time overhead. Motivated by this, we propose a novel Multimodal Federated Learning with Accelerated Knowledge Distillation (MFL-AKD) framework, which is the first attempt to integrate knowledge distillation to combat stragglers in complex multimodal federated scenarios. Concretely, given the pretrained large-scale vision-language models deployed in the central server, we apply a fast knowledge transfer mechanism to conduct early training of local models with part of the local data. The early-trained model is then enhanced through the distillation of the pretrained large model and further trained on the remaining data. Extensive experiments on two datasets for video moment retrieval and two datasets for image-text retrieval demonstrate that our method achieves superior results with high straggler robustness.

Recently developed generative methods, including invertible rescaling network (IRN) based and generative adversarial network (GAN) based methods, have demonstrated exceptional performance in image rescaling. However, IRN-based methods tend to produce over-smoothed results, while GAN-based methods easily generate fake details, which thus hinders their real applications. To address this issue, we propose Boundary-aware Decoupled Flow Networks (BDFlow) to generate realistic and visually pleasing results. Unlike previous methods that model high-frequency information as standard Gaussian distribution directly, our BDFlow first decouples the high-frequency information into semantic high-frequency that adheres to a Boundary distribution and non-semantic high-frequency counterpart that adheres to a Gaussian distribution. Specifically, to capture semantic high-frequency parts accurately, we use Boundary-aware Mask (BAM) to constrain the model to produce rich textures, while non-semantic high-frequency part is randomly sampled from a Gaussian distribution. Comprehensive experiments demonstrate that our BDFlow significantly outperforms other state-of-the-art methods while maintaining lower complexity. Notably, our BDFlow improves the PSNR by 4.4 dB and the SSIM by 0.1 on average over GRAIN, utilizing only 74% of the parameters and 20% of the computation. The code will be available at https://github.com/THU-Kingmin/BAFlow.

Multi-label learning (MLL) requires comprehensive multi-semantic annotations that is hard to fully obtain, thus often resulting in missing labels scenarios. In this paper, we investigate Single Positive Multi-label Learning (SPML), where each image is associated with merely one positive label. Existing SPML methods only focus on designing losses using mechanisms such as hard pseudo-labeling and robust losses, mostly leading to unacceptable false negatives. To address this issue, we first propose a generalized loss framework based on expected risk minimization to provide soft pseudo labels, and point out that the former losses can be seamlessly converted into our framework. In particular, we design a novel robust loss based on our framework, which enjoys flexible coordination between false positives and false negatives, and can additionally deal with the imbalance between positive and negative samples. Extensive experiments show that our approach can significantly improve SPML performance and outperform the vast majority of state-of-the-art methods on all the four benchmarks. Our code is available at https://github.com/yan4xi1/GRLoss.

Data augmentation plays a pivotal role in enhancing and diversifying training data. Nonetheless, consistently improving model performance in varied learning scenarios, especially those with inherent data biases, remains challenging. To address this, we propose to augment the deep features of samples by incorporating their adversarial and anti-adversarial perturbation distributions, enabling adaptive adjustment in the learning difficulty tailored to each sample’s specific characteristics. We then theoretically reveal that our augmentation process approximates the optimization of a surrogate loss function as the number of augmented copies increases indefinitely. This insight leads us to develop a meta-learning-based framework for optimizing classifiers with this novel loss, introducing the effects of augmentation while bypassing the explicit augmentation process. We conduct extensive experiments across four common biased learning scenarios: long-tail learning, generalized long-tail learning, noisy label learning, and subpopulation shift learning. The empirical results demonstrate that our method consistently achieves state-of-the-art performance, highlighting its broad adaptability.

The effectiveness of model training heavily relies on the quality of available training resources. However, budget constraints often impose limitations on data collection efforts. To tackle this challenge, we introduce causal exploration in this paper, a strategy that leverages the underlying causal knowledge for both data collection and model training. We, in particular, focus on enhancing the sample efficiency and reliability of the world model learning within the domain of task-agnostic reinforcement learning. During the exploration phase, the agent actively selects actions expected to yield causal insights most beneficial for world model training. Concurrently, the causal knowledge is acquired and incrementally refined with the ongoing collection of data. We demonstrate that causal exploration aids in learning accurate world models using fewer data and provide theoretical guarantees for its convergence. Empirical experiments, on both synthetic data and real-world applications, further validate the benefits of causal exploration. The source code is available at https://github.com/CMACH508/CausalExploration.

Diffusion probabilistic models (DPMs) have been shown to generate high-quality images without the need for delicate adversarial training. The sampling process of DPMs is mathematically similar to Stochastic Gradient Descent (SGD), with both being iteratively updated with a function increment. Building on this, we present a novel reverse sampler for DPMs in this paper, drawing inspiration from the widely-used Adam optimizer. Our proposed sampler can be readily applied to a pre-trained diffusion model, utilizing momentum mechanisms and adaptive updating to enhance the generated image's quality. By effectively reusing update directions from early steps, our proposed sampler achieves a better balance between high-level semantics and low-level details. Additionally, this sampler is flexible and can be easily integrated into pre-trained DPMs regardless of the sampler used during training. Our experimental results on multiple benchmarks demonstrate that our proposed reverse sampler yields remarkable improvements over different baselines.

Witnessing the evolution of text-to-image diffusion models, significant strides have been made in text-to-3D generation. Currently, two primary paradigms dominate the field of text-to-3D: the feed-forward generation solutions, capable of swiftly producing 3D assets but often yielding coarse results, and the Score Distillation Sampling (SDS) based solutions, known for generating high-fidelity 3D assets albeit at a slower pace. The synergistic integration of these methods holds substantial promise for advancing 3D generation techniques. In this paper, we present BoostDream, a highly efficient plug-and-play 3D refining method designed to transform coarse 3D assets into high-quality. The BoostDream framework comprises three distinct processes: (1) We introduce 3D model distillation that fits differentiable representations from the 3D assets obtained through feed-forward generation. (2) A novel multi-view SDS loss is designed, which utilizes a multi-view aware 2D diffusion model to refine the 3D assets. (3) We propose to use prompt and multi-view consistent normal maps as guidance in refinement. Our extensive experiment is conducted on different differentiable 3D representations, revealing that BoostDream excels in generating high-quality 3D assets rapidly, overcoming the Janus problem compared to conventional SDS-based methods. This breakthrough signifies a substantial advancement in both the efficiency and quality of 3D generation processes.

Block-wise missing data poses significant challenges in real-world data imputation tasks. Compared to scattered missing data, block-wise gaps exacerbate adverse effects on subsequent analytic and machine learning tasks, as the lack of local neighboring elements significantly reduces the interpolation capability and predictive power. However, this issue has not received adequate attention. Most SOTA matrix completion methods appeared less effective, primarily due to overreliance on neighboring elements for predictions. We systematically analyze the issue and propose a novel matrix completion method "BlockEcho" for a more comprehensive solution. This method creatively integrates Matrix Factorization (MF) within Generative Adversarial Networks (GAN) to explicitly retain long-distance inter-element relationships in the original matrix. Besides, we incorporate an additional discriminator for GAN, comparing the generator's intermediate progress with pre-trained MF results to constrain high-order feature distributions. Subsequently, we evaluate BlockEcho on public datasets across three domains. Results demonstrate superior performance over both traditional and SOTA methods when imputing block-wise missing data, especially at higher missing rates. The advantage also holds for scattered missing data at high missing rates. We also contribute on the analyses in providing theoretical justification on the optimality and convergence of fusing MF and GAN for missing block data.

Dynamic adaptation and learning, akin to natural organisms, is crucial for robots operating in real-world scenarios like search and rescue missions. We propose a solution combining intuition from embodied evolution and Bayes theory to promote flexible exploration in foraging tasks. Our investigation focuses on three main areas: 1) leveraging communication and prior knowledge to develop adaptable strategies in agent groups, 2) addressing challenges from sparse rewards or limited data availability, and 3) developing methods for concurrent evaluation and training in a single iteration, filling a current gap in learning-based solutions. Future directions include exploring decentralized coordination among agents and incorporating assistance based on prospective memory and altruism in multi-agent reinforcement learning.

Hand gesture recognition (HGR) plays a pivotal role in natural and intuitive human-computer interactions. Recent HGR methods focus on recognizing gestures from vision-based images or videos. However, vision-based methods are limited in recognizing micro hand gestures (MHGs) (e.g., pinch within 1cm) and gestures with occluded fingers. To address these issues, combined with the electromyography (EMG) technique, we propose BeyondVision, an EMG-driven MHG recognition system based on deep learning. BeyondVision consists of a wristband-style EMG sampling device and a tailored lightweight neural network BV-Net that can accurately translate EMG signals of MHGs to control commands in real-time. Moreover, we propose a post-processing mechanism and a weight segmentation algorithm to effectively improve the accuracy rate of MHG recognition. Subjective and objective experimental results show that our approach achieves over 95% average recognition rate, 2000Hz sampling frequency, and real-time micro gesture recognition. Our technique has been applied in a commercially available product, introduced at: https://github.com/tyc333/NoBarriers.

Recently, large language models (LLMs) have shown remarkable capabilities including understanding context, engaging in logical reasoning, and generating responses. However, this is achieved at the expense of stringent computational and memory requirements, hindering their ability to effectively support long input sequences. This survey provides an inclusive review of the recent techniques and methods devised to extend the sequence length in LLMs, thereby enhancing their capacity for long-context understanding. In particular, we review and categorize a wide range of techniques including architectural modifications, such as modified positional encoding and altered attention mechanisms, which are designed to enhance the processing of longer sequences while avoiding a proportional increase in computational cost. The diverse methodologies investigated in this study can be leveraged across different phases of LLMs, i.e., training, fine-tuning and inference. This enables LLMs to efficiently process extended sequences. The limitations of the current methodologies is discussed in the last section along with the suggestions for future research directions, underscoring the importance of sequence length in the continued advancement of LLMs.

Exploring the realms of counterfactuals, this paper introduces a versatile approach in text generation using structural causal models (SCM), broadening the scope beyond traditional singular causal studies to encompass complex, multi-layered relationships. To comprehensively explore these intricate, multi-layered causal relationships in text generation, we introduce a generalized approach based on the structural causal model (SCM), adept at handling complex causal interactions in a spectrum ranging from everyday stories to financial reports.Specifically, our method begins by disentangling each component of the text into pairs of latent variables, representing elements that remain unchanged and those subject to variation. Subsequently, counterfactual interventions are applied to these latent variables, facilitating the generation of outcomes that are influenced by complex causal dynamics. Extensive experiments have been conducted on both a public story generation dataset and a specially constructed dataset in the financial domain. The experimental results demonstrate that our approach achieves state-of-the-art performance across a range of automatic and human evaluation criteria, underscoring its effectiveness and versatility in diverse text generation contexts.

Multiple complex degradations are coupled in low-quality video faces in the real world. Therefore, blind video face restoration is a highly challenging ill-posed problem, requiring not only hallucinating high-fidelity details but also enhancing temporal coherence across diverse pose variations. Restoring each frame independently in a naive manner inevitably introduces temporal incoherence and artifacts from pose changes and keypoint localization errors. To address this, we propose the first blind video face restoration approach with a novel parsing-guided temporal-coherent transformer (PGTFormer) without pre-alignment. PGTFormer leverages semantic parsing guidance to select optimal face priors for generating temporally coherent artifact-free results. Specifically, we pre-train a temporal-spatial vector quantized auto-encoder on high-quality video face datasets to extract expressive context-rich priors. Then, the temporal parse-guided codebook predictor (TPCP) restores faces in different poses based on face parsing context cues without performing face pre-alignment. This strategy reduces artifacts and mitigates jitter caused by cumulative errors from face pre-alignment. Finally, the temporal fidelity regulator (TFR) enhances fidelity through temporal feature interaction and improves video temporal consistency. Extensive experiments on face videos show that our method outperforms previous face restoration baselines. The code will be released on https://github.com/kepengxu/PGTFormer.

Hyperparameter optimization (HPO) is indispensable for achieving optimal performance in machine learning tasks. A popular class of methods in this regard is based on Successive Halving (SHA), which casts HPO into a pure-exploration multi-armed bandit problem under finite sampling budget constraints. This is accomplished by considering hyperparameter configurations as arms and rewards as the negative validation losses. While enjoying theoretical guarantees as well as working well in practice, SHA comes, however, with several hyperparameters itself, one of which is the maximum budget that can be allocated to evaluate a single arm (hyperparameter configuration). Although there are already solutions to this meta hyperparameter optimization problem, such as the doubling trick or asynchronous extensions of SHA, these are either practically inefficient or lack theoretical guarantees. In this paper, we propose incremental SHA (iSHA), a synchronous extension of SHA, allowing to increase the maximum budget a posteriori while still enjoying theoretical guarantees. Our empirical analysis of HPO problems corroborates our theoretical findings and shows that iSHA is more resource-efficient than existing SHA-based approaches.

In recent years, badminton analytics has drawn attention due to the advancement of artificial intelligence and the efficiency of data collection. While there is a line of effective applications to improve and investigate player performance, there are only a few public badminton datasets that can be used by researchers outside the badminton domain. Existing badminton singles datasets focus on specific matchups; however, they cannot provide comprehensive studies on different players and various matchups. In this paper, we provide a badminton singles dataset, ShuttleSet22, which is collected from high-ranking matches in 2022. ShuttleSet22 consists of 30,172 strokes in 2,888 rallies in the training set, 1,400 strokes in 450 rallies in the validation set, and 2,040 strokes in 654 rallies in the testing set, with detailed stroke-level metadata within a rally. To benchmark existing work with ShuttleSet22, we hold a challenge, Track 2: Forecasting Future Turn-Based Strokes in Badminton Rallies, at CoachAI Badminton Challenge @ IJCAI 2023, to encourage researchers to tackle this real-world problem through innovative approaches and to summarize insights between the state-of-the-art baseline and improved techniques, exchanging inspiring ideas. The baseline codes and the dataset are made available at https://github.com/wywyWang/CoachAI-Projects/tree/main/CoachAI-Challenge-IJCAI2023.