The advent of 3D Gaussian Splatting (3DGS) has recently brought about a revolution in the field of neural rendering, facilitating high-quality renderings at real-time speed. However, 3DGS heavily depends on the initialized point cloud produced by Structure-from-Motion (SfM) techniques. When tackling with large-scale scenes that unavoidably contain texture-less surfaces, the SfM techniques always fail to produce enough points in these surfaces and cannot provide good initialization for 3DGS. As a result, 3DGS suffers from difficult optimization and low-quality renderings. In this paper, inspired by classical multi-view stereo (MVS) techniques, we propose GaussianPro, a novel method that applies a progressive propagation strategy to guide the densification of the 3D Gaussians. Compared to the simple split and clone strategies used in 3DGS, our method leverages the priors of the existing reconstructed geometries of the scene and patch matching techniques to produce new Gaussians with accurate positions and orientations. Experiments on both large-scale and small-scale scenes validate the effectiveness of our method, where our method significantly surpasses 3DGS on the Waymo dataset, exhibiting an improvement of 1.15dB in terms of PSNR.
Graph pooling compresses graph information into a compact representation. State-of-the-art graph pooling methods follow a hierarchical approach, which reduces the graph size step-by-step. These methods must balance memory efficiency with preserving node information, depending on whether they use node dropping or node clustering. Additionally, fixed pooling ratios or numbers of pooling layers are predefined for all graphs, which prevents personalized pooling structures from being captured for each individual graph. In this work, inspired by bottom-up grammar induction, we propose an efficient graph parsing algorithm to infer the pooling structure, which then drives graph pooling. The resulting Graph Parsing Network (GPN) adaptively learns personalized pooling structure for each individual graph. GPN benefits from the discrete assignments generated by the graph parsing algorithm, allowing good memory efficiency while preserving node information intact. Experimental results on standard benchmarks demonstrate that GPN outperforms state-of-the-art graph pooling methods in graph classification tasks while being able to achieve competitive performance in node classification tasks. We also conduct a graph reconstruction task to show GPN's ability to preserve node information and measure both memory and time efficiency through relevant tests.
In human activity recognition (HAR), the availability of substantial ground truth is necessary for training efficient models. However, acquiring ground pressure data through physical sensors itself can be cost-prohibitive, time-consuming. To address this critical need, we introduce Text-to-Pressure (T2P), a framework designed to generate extensive ground pressure sequences from textual descriptions of human activities using deep learning techniques. We show that the combination of vector quantization of sensor data along with simple text conditioned auto regressive strategy allows us to obtain high-quality generated pressure sequences from textual descriptions with the help of discrete latent correlation between text and pressure maps. We achieved comparable performance on the consistency between text and generated motion with an R squared value of 0.722, Masked R squared value of 0.892, and FID score of 1.83. Additionally, we trained a HAR model with the the synthesized data and evaluated it on pressure dynamics collected by a real pressure sensor which is on par with a model trained on only real data. Combining both real and synthesized training data increases the overall macro F1 score by 5.9 percent.
Many deep learning synthetic speech generation tools are readily available. The use of synthetic speech has caused financial fraud, impersonation of people, and misinformation to spread. For this reason forensic methods that can detect synthetic speech have been proposed. Existing methods often overfit on one dataset and their performance reduces substantially in practical scenarios such as detecting synthetic speech shared on social platforms. In this paper we propose, Patched Spectrogram Synthetic Speech Detection Transformer (PS3DT), a synthetic speech detector that converts a time domain speech signal to a mel-spectrogram and processes it in patches using a transformer neural network. We evaluate the detection performance of PS3DT on ASVspoof2019 dataset. Our experiments show that PS3DT performs well on ASVspoof2019 dataset compared to other approaches using spectrogram for synthetic speech detection. We also investigate generalization performance of PS3DT on In-the-Wild dataset. PS3DT generalizes well than several existing methods on detecting synthetic speech from an out-of-distribution dataset. We also evaluate robustness of PS3DT to detect telephone quality synthetic speech and synthetic speech shared on social platforms (compressed speech). PS3DT is robust to compression and can detect telephone quality synthetic speech better than several existing methods.
Tuberculosis (TB) remains a significant global health challenge, with pediatric cases posing a major concern. The World Health Organization (WHO) advocates for chest X-rays (CXRs) for TB screening. However, visual interpretation by radiologists can be subjective, time-consuming and prone to error, especially in pediatric TB. Artificial intelligence (AI)-driven computer-aided detection (CAD) tools, especially those utilizing deep learning, show promise in enhancing lung disease detection. However, challenges include data scarcity and lack of generalizability. In this context, we propose a novel self-supervised paradigm leveraging Vision Transformers (ViT) for improved TB detection in CXR, enabling zero-shot pediatric TB detection. We demonstrate improvements in TB detection performance ($\sim$12.7% and $\sim$13.4% top AUC/AUPR gains in adults and children, respectively) when conducting self-supervised pre-training when compared to fully-supervised (i.e., non pre-trained) ViT models, achieving top performances of 0.959 AUC and 0.962 AUPR in adult TB detection, and 0.697 AUC and 0.607 AUPR in zero-shot pediatric TB detection. As a result, this work demonstrates that self-supervised learning on adult CXRs effectively extends to challenging downstream tasks such as pediatric TB detection, where data are scarce.
Single photon detection and timing gathered increasing interest in the last few years due to both its necessity in the field of quantum sensing and the advantages of single quanta detection in the field of low level light imaging. While simple bucket detectors are mature enough for commercial applications, more complex imaging detectors are still a field of research with mostly prototype level detectors. A major problem in these detectors is the implementation of in-pixel timing circuitry, especially for two-dimensional imagers. One of the most promising approaches is the use of voltage controlled ring resonators in every pixel. Each of those is running independently, based on a voltage supplied by a global reference. However, this yields the problem that across the chip the supply voltage can change, which in turn changes the period of the ring resonator. Due to additional parasitic effects, this problem can worsen with increasing measurement time, leading to a drift of the timing information. We present here a method to identify and correct such temporal drifts of single photon detectors, based on asynchronous quantum ghost imaging. We also show the effect of this correction on a recent QGI measurement from our group.
Coordinate-based neural implicit representation or implicit fields have been widely studied for 3D geometry representation or novel view synthesis. Recently, a series of efforts have been devoted to accelerating the speed and improving the quality of the coordinate-based implicit field learning. Instead of learning heavy MLPs to predict the neural implicit values for the query coordinates, neural voxels or grids combined with shallow MLPs have been proposed to achieve high-quality implicit field learning with reduced optimization time. On the other hand, lightweight field representations such as linear grid have been proposed to further improve the learning speed. In this paper, we aim for both fast and high-quality implicit field learning, and propose TaylorGrid, a novel implicit field representation which can be efficiently computed via direct Taylor expansion optimization on 2D or 3D grids. As a general representation, TaylorGrid can be adapted to different implicit fields learning tasks such as SDF learning or NeRF. From extensive quantitative and qualitative comparisons, TaylorGrid achieves a balance between the linear grid and neural voxels, showing its superiority in fast and high-quality implicit field learning.
Convolutional neural networks (CNNs) with large kernels, drawing inspiration from the key operations of vision transformers (ViTs), have demonstrated impressive performance in various vision-based applications. To address the issue of computational efficiency degradation in existing designs for supporting large-kernel convolutions, an FPGA-based inference accelerator is proposed for the efficient deployment of CNNs with arbitrary kernel sizes. Firstly, a Z-flow method is presented to optimize the computing data flow by maximizing data reuse opportunity. Besides, the proposed design, incorporating the kernel-segmentation (Kseg) scheme, enables extended support for large-kernel convolutions, significantly reducing the storage requirements for overlapped data. Moreover, based on the analysis of typical block structures in emerging CNNs, vertical-fused (VF) and horizontal-fused (HF) methods are developed to optimize CNN deployments from both computation and transmission perspectives. The proposed hardware accelerator, evaluated on Intel Arria 10 FPGA, achieves up to 3.91 times better DSP efficiency than prior art on the same network. Particularly, it demonstrates efficient support for large-kernel CNNs, achieving throughputs of 169.68 GOPS and 244.55 GOPS for RepLKNet-31 and PyConvResNet-50, respectively, both of which are implemented on hardware for the first time.
Recent methods have demonstrated that Large Language Models (LLMs) can solve reasoning tasks better when they are encouraged to solve subtasks of the main task first. In this paper we devise a similar strategy that breaks down reasoning tasks into a problem decomposition phase and a problem solving phase and show that the strategy is able to outperform a single stage solution. Further, we hypothesize that the decomposition should be easier to distill into a smaller model compared to the problem solving because the latter requires large amounts of domain knowledge while the former only requires learning general problem solving strategies. We propose methods to distill these two capabilities and evaluate their impact on reasoning outcomes and inference cost. We find that we can distill the problem decomposition phase and at the same time achieve good generalization across tasks, datasets, and models. However, it is harder to distill the problem solving capability without losing performance and the resulting distilled model struggles with generalization. These results indicate that by using smaller, distilled problem decomposition models in combination with problem solving LLMs we can achieve reasoning with cost-efficient inference and local adaptation.
The recent increase in renewable energy penetration at the distribution level introduces a multi-directional power flow that outdated traditional fault location techniques. To this extent, the development of new methods is needed to ensure fast and accurate fault localization and, hence, strengthen power system reliability. This paper proposes a data-driven ground fault location method for the power distribution system. An 11-bus 20 kV power system is modeled in Matlab/Simulink to simulate ground faults. The faults are generated at different locations and under various system operational states. Time-domain faulted three-phase voltages at the system substation are then analyzed with discrete wavelet transform. Statistical quantities of the processed data are eventually used to train an Artificial Neural Network (ANN) to find a mapping between computed voltage features and faults. Specifically, three ANNs allow the prediction of faulted phase, faulted branch, and fault distance from the system substation separately. According to the results, the method shows good potential, with a total relative error of 0,4% for fault distance prediction. The method is applied to datasets with unknown system states to test robustness.