arXiv subjects

Peng Jia

Publications and source records attributed to Peng Jia.

At least 55 records · Page 3Linked to original sources

arXiv2025

TokenFLEX: Unified VLM Training for Flexible Visual Tokens Inference

Conventional Vision-Language Models(VLMs) typically utilize a fixed number of vision tokens, regardless of task complexity. This one-size-fits-all strategy introduces notable inefficiencies: using excessive tokens leads to unnecessary computational overhead in simpler tasks, whereas insufficient tokens compromise fine-grained visual comprehension in more complex contexts. To overcome these limitations, we present TokenFLEX, an innovative and adaptable vision-language framework that encodes images into a variable number of tokens for efficient integration with a Large Language Model (LLM). Our approach is underpinned by two pivotal innovations. Firstly, we present a novel training paradigm that enhances performance across varying numbers of vision tokens by stochastically modulating token counts during training. Secondly, we design a lightweight vision token projector incorporating an adaptive pooling layer and SwiGLU, allowing for flexible downsampling of vision tokens and adaptive selection of features tailored to specific token counts. Comprehensive experiments reveal that TokenFLEX consistently outperforms its fixed-token counterparts, achieving notable performance gains across various token counts enhancements of 1.6%, 1.0%, and 0.4% with 64, 144, and 256 tokens, respectively averaged over eight vision-language benchmarks. These results underscore TokenFLEX's remarkable flexibility while maintaining high-performance vision-language understanding.

cs.CV↗

arXiv2025

VISTA: Unsupervised 2D Temporal Dependency Representations for Time Series Anomaly Detection

Time Series Anomaly Detection (TSAD) is essential for uncovering rare and potentially harmful events in unlabeled time series data. Existing methods are highly dependent on clean, high-quality inputs, making them susceptible to noise and real-world imperfections. Additionally, intricate temporal relationships in time series data are often inadequately captured in traditional 1D representations, leading to suboptimal modeling of dependencies. We introduce VISTA, a training-free, unsupervised TSAD algorithm designed to overcome these challenges. VISTA features three core modules: 1) Time Series Decomposition using Seasonal and Trend Decomposition via Loess (STL) to decompose noisy time series into trend, seasonal, and residual components; 2) Temporal Self-Attention, which transforms 1D time series into 2D temporal correlation matrices for richer dependency modeling and anomaly detection; and 3) Multivariate Temporal Aggregation, which uses a pretrained feature extractor to integrate cross-variable information into a unified, memory-efficient representation. VISTA's training-free approach enables rapid deployment and easy hyperparameter tuning, making it suitable for industrial applications. It achieves state-of-the-art performance on five multivariate TSAD benchmarks.

cs.LG↗

arXiv2025

StyledStreets: Multi-style Street Simulator with Spatial and Temporal Consistency

Urban scene reconstruction requires modeling both static infrastructure and dynamic elements while supporting diverse environmental conditions. We present \textbf{StyledStreets}, a multi-style street simulator that achieves instruction-driven scene editing with guaranteed spatial and temporal consistency. Building on a state-of-the-art Gaussian Splatting framework for street scenarios enhanced by our proposed pose optimization and multi-view training, our method enables photorealistic style transfers across seasons, weather conditions, and camera setups through three key innovations: First, a hybrid embedding scheme disentangles persistent scene geometry from transient style attributes, allowing realistic environmental edits while preserving structural integrity. Second, uncertainty-aware rendering mitigates supervision noise from diffusion priors, enabling robust training across extreme style variations. Third, a unified parametric model prevents geometric drift through regularized updates, maintaining multi-view consistency across seven vehicle-mounted cameras. Our framework preserves the original scene's motion patterns and geometric relationships. Qualitative results demonstrate plausible transitions between diverse conditions (snow, sandstorm, night), while quantitative evaluations show state-of-the-art geometric accuracy under style transfers. The approach establishes new capabilities for urban simulation, with applications in autonomous vehicle testing and augmented reality systems requiring reliable environmental consistency. Codes will be publicly available upon publication.

cs.CV↗

arXiv2025

Image Pre-Processing Framework for Time-Domain Astronomy in the Artificial Intelligence Era

The rapid advancement of image analysis methods in time-domain astronomy, particularly those leveraging AI algorithms, has highlighted efficient image pre-processing as a critical bottleneck affecting algorithm performance. Image pre-processing, which involves standardizing images for training or deployment of various AI algorithms, encompasses essential steps such as image quality evaluation, alignment, stacking, background extraction, gray-scale transformation, cropping, source detection, astrometry, and photometry. Historically, these algorithms were developed independently by different research groups, primarily based on CPU architecture for small-scale data processing. This paper introduces a novel framework for image pre-processing that integrates key algorithms specifically modified for GPU architecture, enabling large-scale image pre-processing for different algorithms. To prepare for the new algorithm design paradigm in the AI era, we have implemented two operational modes in the framework for different application scenarios: Eager mode and Pipeline mode. The Eager mode facilitates real-time feedback and flexible adjustments, which could be used for parameter tuning and algorithm development. The pipeline mode is primarily designed for large scale data processing, which could be used for training or deploying of artificial intelligence models. We have tested the performance of our framework using simulated and real observation images. Results demonstrate that our framework significantly enhances image pre-processing speed while maintaining accuracy levels comparable to CPU based algorithms. To promote accessibility and ease of use, a Docker version of our framework is available for download in the PaperData Repository powered by China-VO, compatible with various AI algorithms developed for time-domain astronomy research.

astro-ph.IM↗

arXiv2025

An Ultra-Fast Image Simulation Technique with Spatially Variable Point Spread Functions

Simulated images are essential in algorithm development and instrument testing for optical telescopes. During real observations, images obtained by optical telescopes are affected by spatially variable point spread functions (PSFs), a crucial effect requiring accurate simulation. Traditional methods segment images into patches, convolve patches with individual PSFs, and reassemble them as a whole image. Although widely used, these approaches suffer from slow convolution processes and reduced image fidelity due to abrupt PSF transitions between different patches. This paper introduces a novel method for generating simulated images with spatial continuously varying PSFs. Our approach firstly decomposes original images into PSF bases derived with the principal component analysis method. The entire image is then convolved with these PSF bases to create image bases. Finally, we multiply the coefficients of image bases with these image bases for each pixels and add the multiplication results along each pixel to obtain the final simulated image. Our method could generate high-fidelity simulated images with spatially variable PSFs without boundary artifacts. The method proposed in this paper significantly improves the speed of astronomical image simulation, potentially advancing observational astronomy and instrumental development.

astro-ph.IM↗

arXiv2024

GaussianAD: Gaussian-Centric End-to-End Autonomous Driving

Vision-based autonomous driving shows great potential due to its satisfactory performance and low costs. Most existing methods adopt dense representations (e.g., bird's eye view) or sparse representations (e.g., instance boxes) for decision-making, which suffer from the trade-off between comprehensiveness and efficiency. This paper explores a Gaussian-centric end-to-end autonomous driving (GaussianAD) framework and exploits 3D semantic Gaussians to extensively yet sparsely describe the scene. We initialize the scene with uniform 3D Gaussians and use surrounding-view images to progressively refine them to obtain the 3D Gaussian scene representation. We then use sparse convolutions to efficiently perform 3D perception (e.g., 3D detection, semantic map construction). We predict 3D flows for the Gaussians with dynamic semantics and plan the ego trajectory accordingly with an objective of future scene forecasting. Our GaussianAD can be trained in an end-to-end manner with optional perception labels when available. Extensive experiments on the widely used nuScenes dataset verify the effectiveness of our end-to-end GaussianAD on various tasks including motion planning, 3D occupancy prediction, and 4D occupancy forecasting. Code: https://github.com/wzzheng/GaussianAD.

cs.CV↗

arXiv2024

ReconDreamer: Crafting World Models for Driving Scene Reconstruction via Online Restoration

Closed-loop simulation is crucial for end-to-end autonomous driving. Existing sensor simulation methods (e.g., NeRF and 3DGS) reconstruct driving scenes based on conditions that closely mirror training data distributions. However, these methods struggle with rendering novel trajectories, such as lane changes. Recent works have demonstrated that integrating world model knowledge alleviates these issues. Despite their efficiency, these approaches still encounter difficulties in the accurate representation of more complex maneuvers, with multi-lane shifts being a notable example. Therefore, we introduce ReconDreamer, which enhances driving scene reconstruction through incremental integration of world model knowledge. Specifically, DriveRestorer is proposed to mitigate artifacts via online restoration. This is complemented by a progressive data update strategy designed to ensure high-quality rendering for more complex maneuvers. To the best of our knowledge, ReconDreamer is the first method to effectively render in large maneuvers. Experimental results demonstrate that ReconDreamer outperforms Street Gaussians in the NTA-IoU, NTL-IoU, and FID, with relative improvements by 24.87%, 6.72%, and 29.97%. Furthermore, ReconDreamer surpasses DriveDreamer4D with PVG during large maneuver rendering, as verified by a relative improvement of 195.87% in the NTA-IoU metric and a comprehensive user study.

cs.CV↗

arXiv2024

UPdec-Webb: A Dataset for Coaddition of JWST NIRCam Images

We present the application of the image coaddition algorithm, Up-sampling and PSF Deconvolution Coaddition (UPDC), for stacking multiple exposure images captured by the James Webb Space Telescope (JWST) Near-Infrared Camera (NIRCam). By addressing the point spread function (PSF) effect, UPDC provides visually enhanced and sharper images. Furthermore, the anti-aliasing and super-resolution capabilities of UPDC make it easier to deblend sources overlapped on images, yielding a higher accuracy of aperture photometry. We apply this algorithm to the SMACS J0723 imaging data. Comparative analysis with the Drizzle algorithm demonstrates significant improvements in detecting faint sources, achieving accurate photometry, and effectively deblending (super-resolution) closely packed sources. {As a result, we have newly detected a pair of close binary stars that were previously unresolvable in the original exposures or the Drizzled image.} These improvements significantly benefit various scientific projects conducted by JWST. The resulting dataset, named "UPdec-Webb", can be accessible through the official website of the Chinese Virtual Observatory (ChinaVO).

astro-ph.IM↗

arXiv2024

Generalizing Motion Planners with Mixture of Experts for Autonomous Driving

Large real-world driving datasets have sparked significant research into various aspects of data-driven motion planners for autonomous driving. These include data augmentation, model architecture, reward design, training strategies, and planner pipelines. These planners promise better generalizations on complicated and few-shot cases than previous methods. However, experiment results show that many of these approaches produce limited generalization abilities in planning performance due to overly complex designs or training paradigms. In this paper, we review and benchmark previous methods focusing on generalizations. The experimental results indicate that as models are appropriately scaled, many design elements become redundant. We introduce StateTransformer-2 (STR2), a scalable, decoder-only motion planner that uses a Vision Transformer (ViT) encoder and a mixture-of-experts (MoE) causal Transformer architecture. The MoE backbone addresses modality collapse and reward balancing by expert routing during training. Extensive experiments on the NuPlan dataset show that our method generalizes better than previous approaches across different test sets and closed-loop simulations. Furthermore, we assess its scalability on billions of real-world urban driving scenarios, demonstrating consistent accuracy improvements as both data and model size grow.

cs.RO↗

arXiv2024

DAT: Dialogue-Aware Transformer with Modality-Group Fusion for Human Engagement Estimation

Engagement estimation plays a crucial role in understanding human social behaviors, attracting increasing research interests in fields such as affective computing and human-computer interaction. In this paper, we propose a Dialogue-Aware Transformer framework (DAT) with Modality-Group Fusion (MGF), which relies solely on audio-visual input and is language-independent, for estimating human engagement in conversations. Specifically, our method employs a modality-group fusion strategy that independently fuses audio and visual features within each modality for each person before inferring the entire audio-visual content. This strategy significantly enhances the model's performance and robustness. Additionally, to better estimate the target participant's engagement levels, the introduced Dialogue-Aware Transformer considers both the participant's behavior and cues from their conversational partners. Our method was rigorously tested in the Multi-Domain Engagement Estimation Challenge held by MultiMediate'24, demonstrating notable improvements in engagement-level regression precision over the baseline model. Notably, our approach achieves a CCC score of 0.76 on the NoXi Base test set and an average CCC of 0.64 across the NoXi Base, NoXi-Add, and MPIIGI test sets.

cs.HC↗

arXiv2024

DiVE: DiT-based Video Generation with Enhanced Control

Generating high-fidelity, temporally consistent videos in autonomous driving scenarios faces a significant challenge, e.g. problematic maneuvers in corner cases. Despite recent video generation works are proposed to tackcle the mentioned problem, i.e. models built on top of Diffusion Transformers (DiT), works are still missing which are targeted on exploring the potential for multi-view videos generation scenarios. Noticeably, we propose the first DiT-based framework specifically designed for generating temporally and multi-view consistent videos which precisely match the given bird's-eye view layouts control. Specifically, the proposed framework leverages a parameter-free spatial view-inflated attention mechanism to guarantee the cross-view consistency, where joint cross-attention modules and ControlNet-Transformer are integrated to further improve the precision of control. To demonstrate our advantages, we extensively investigate the qualitative comparisons on nuScenes dataset, particularly in some most challenging corner cases. In summary, the effectiveness of our proposed method in producing long, controllable, and highly consistent videos under difficult conditions is proven to be effective.

cs.CV↗

arXiv2024

First Discovery and Confirmation of PN Candidates Found from AI and Deep Learning Techniques Applied to VPHAS+ Survey Data

Context. We have developed deep learning (DL) and AI-based tools to search extant narrow-band wide-field H$α$ surveys of the Galactic Plane for elusive planetary nebulae (PNe) which are hidden in dense star fields towards the Galactic center. They are faint, low-surface brightness, usually resolved sources, which are not discovered by previous automatic searches that depend on photometric data for point-like sources. These sources are very challenging to find by traditional visual inspection in such crowded fields and many have been missed. We have successfully adopted a novel 'Swin-Transformer' AI algorithm, which we described in detail in the preceding Techniques paper (Paper I). Aims. Here, we present preliminary results from our first spectroscopic follow-up run for 31 top-quality PN candidates found by the algorithm from the high-resolution H$α$ survey VPHAS+. This survey has not yet undergone extensive manual, systematic searching. Methods. Our candidate PNe were observed with the SpUpNIC spectrograph on the 1.9 m telescope at the South African Astronomical Observatory (SAAO) in June 2023. We performed standard IRAF spectroscopic reduction and then followed our normal HASH PN identification and classification procedures. Results. Our reduced spectra confirmed that these candidates include 22 true, likely, and possible PNe (70.97\%), 3 emission-line galaxies, 2 emission-line stars, 2 late-type star contaminants, and 2 other H$α$ sources including a newly identified detached fragment of SNR RCW 84. We present the imaging and spectral data of these candidates and a preliminary analysis of their properties. These data provide strong input to help evaluate and refine the behavior of the AI algorithm when searching for PNe in wide-field H$α$ surveys.

astro-ph.GA↗

arXiv2024

DriveVLM: The Convergence of Autonomous Driving and Large Vision-Language Models

A primary hurdle of autonomous driving in urban environments is understanding complex and long-tail scenarios, such as challenging road conditions and delicate human behaviors. We introduce DriveVLM, an autonomous driving system leveraging Vision-Language Models (VLMs) for enhanced scene understanding and planning capabilities. DriveVLM integrates a unique combination of reasoning modules for scene description, scene analysis, and hierarchical planning. Furthermore, recognizing the limitations of VLMs in spatial reasoning and heavy computational requirements, we propose DriveVLM-Dual, a hybrid system that synergizes the strengths of DriveVLM with the traditional autonomous driving pipeline. Experiments on both the nuScenes dataset and our SUP-AD dataset demonstrate the efficacy of DriveVLM and DriveVLM-Dual in handling complex and unpredictable driving conditions. Finally, we deploy the DriveVLM-Dual on a production vehicle, verifying it is effective in real-world autonomous driving environments.

cs.CV↗

arXiv2024

BEV-TSR: Text-Scene Retrieval in BEV Space for Autonomous Driving

The rapid development of the autonomous driving industry has led to a significant accumulation of autonomous driving data. Consequently, there comes a growing demand for retrieving data to provide specialized optimization. However, directly applying previous image retrieval methods faces several challenges, such as the lack of global feature representation and inadequate text retrieval ability for complex driving scenes. To address these issues, firstly, we propose the BEV-TSR framework which leverages descriptive text as an input to retrieve corresponding scenes in the Bird's Eye View (BEV) space. Then to facilitate complex scene retrieval with extensive text descriptions, we employ a large language model (LLM) to extract the semantic features of the text inputs and incorporate knowledge graph embeddings to enhance the semantic richness of the language embedding. To achieve feature alignment between the BEV feature and language embedding, we propose Shared Cross-modal Embedding with a set of shared learnable embeddings to bridge the gap between these two modalities, and employ a caption generation task to further enhance the alignment. Furthermore, there lack of well-formed retrieval datasets for effective evaluation. To this end, we establish a multi-level retrieval dataset, nuScenes-Retrieval, based on the widely adopted nuScenes dataset. Experimental results on the multi-level nuScenes-Retrieval show that BEV-TSR achieves state-of-the-art performance, e.g., 85.78% and 87.66% top-1 accuracy on scene-to-text and text-to-scene retrieval respectively. Codes and datasets will be available.

cs.CV↗

arXiv2024

RMFA-Net: A Neural ISP for Real RAW to RGB Image Reconstruction

Deep learning-based ISP algorithms have demonstrated significant potential in raw2rgb reconstruction. However, existing networks have not fully considered the specific characteristics of raw data, such as black level and CFA, which can negatively impact texture and color if mishandled. Moreover, uneven exposure in raw data is also not considered carefully, leading to adverse effects on contrast and brightness. In this paper, we introduce RMFA-Net to tackle these problems. We perform implicit black level correction to mitigate color shifts in dim scenes. To preserve high-frequency information and prevent misalignment, we propose a novel Three-Channel-Split mode. To address the issue of uneven exposure, we designed an explicit tone mapping module based on the Retinex theory. We train and evaluate our models using the dataset released by the Mobile AI 2022 Learned Smartphone ISP Challenge. It is demonstrated that RMFA-Net outperforms previous algorithms, achieving a PSNR score of over 25 dB, surpassing the state-of-the-art by +1 dB. Furthermore, we developed a lightweight version, RMFANet-tiny, for engineering deployment while still maintaining strong performance, surpassing the SOTA by +0.5 dB.

eess.IV↗

arXiv2024

Unleashing Generalization of End-to-End Autonomous Driving with Controllable Long Video Generation

Using generative models to synthesize new data has become a de-facto standard in autonomous driving to address the data scarcity issue. Though existing approaches are able to boost perception models, we discover that these approaches fail to improve the performance of planning of end-to-end autonomous driving models as the generated videos are usually less than 8 frames and the spatial and temporal inconsistencies are not negligible. To this end, we propose Delphi, a novel diffusion-based long video generation method with a shared noise modeling mechanism across the multi-views to increase spatial consistency, and a feature-aligned module to achieves both precise controllability and temporal consistency. Our method can generate up to 40 frames of video without loss of consistency which is about 5 times longer compared with state-of-the-art methods. Instead of randomly generating new data, we further design a sampling policy to let Delphi generate new data that are similar to those failure cases to improve the sample efficiency. This is achieved by building a failure-case driven framework with the help of pre-trained visual language models. Our extensive experiment demonstrates that our Delphi generates a higher quality of long videos surpassing previous state-of-the-art methods. Consequentially, with only generating 4% of the training dataset size, our framework is able to go beyond perception and prediction tasks, for the first time to the best of our knowledge, boost the planning performance of the end-to-end autonomous driving model by a margin of 25%.

cs.CV↗

arXiv2024

TOD3Cap: Towards 3D Dense Captioning in Outdoor Scenes

3D dense captioning stands as a cornerstone in achieving a comprehensive understanding of 3D scenes through natural language. It has recently witnessed remarkable achievements, particularly in indoor settings. However, the exploration of 3D dense captioning in outdoor scenes is hindered by two major challenges: 1) the domain gap between indoor and outdoor scenes, such as dynamics and sparse visual inputs, makes it difficult to directly adapt existing indoor methods; 2) the lack of data with comprehensive box-caption pair annotations specifically tailored for outdoor scenes. To this end, we introduce the new task of outdoor 3D dense captioning. As input, we assume a LiDAR point cloud and a set of RGB images captured by the panoramic camera rig. The expected output is a set of object boxes with captions. To tackle this task, we propose the TOD3Cap network, which leverages the BEV representation to generate object box proposals and integrates Relation Q-Former with LLaMA-Adapter to generate rich captions for these objects. We also introduce the TOD3Cap dataset, the largest one to our knowledge for 3D dense captioning in outdoor scenes, which contains 2.3M descriptions of 64.3K outdoor objects from 850 scenes. Notably, our TOD3Cap network can effectively localize and caption 3D objects in outdoor scenes, which outperforms baseline methods by a significant margin (+9.6 CiDEr@0.5IoU). Code, data, and models are publicly available at https://github.com/jxbbb/TOD3Cap.

cs.CV↗

arXiv2024

A Versatile Framework for Analyzing Galaxy Image Data by Implanting Human-in-the-loop on a Large Vision Model

The exponential growth of astronomical datasets provides an unprecedented opportunity for humans to gain insight into the Universe. However, effectively analyzing this vast amount of data poses a significant challenge. Astronomers are turning to deep learning techniques to address this, but the methods are limited by their specific training sets, leading to considerable duplicate workloads too. Hence, as an example to present how to overcome the issue, we built a framework for general analysis of galaxy images, based on a large vision model (LVM) plus downstream tasks (DST), including galaxy morphological classification, image restoration, object detection, parameter extraction, and more. Considering the low signal-to-noise ratio of galaxy images and the imbalanced distribution of galaxy categories, we have incorporated a Human-in-the-loop (HITL) module into our large vision model, which leverages human knowledge to enhance the reliability and interpretability of processing galaxy images interactively. The proposed framework exhibits notable few-shot learning capabilities and versatile adaptability to all the abovementioned tasks on galaxy images in the DESI legacy imaging surveys. Expressly, for object detection, trained by 1000 data points, our DST upon the LVM achieves an accuracy of 96.7%, while ResNet50 plus Mask R-CNN gives an accuracy of 93.1%; for morphology classification, to obtain AUC ~0.9, LVM plus DST and HITL only requests 1/50 training sets compared to ResNet18. Expectedly, multimodal data can be integrated similarly, which opens up possibilities for conducting joint analyses with datasets spanning diverse domains in the era of multi-message astronomy.

astro-ph.IM↗