The study of non-line-of-sight (NLOS) imaging is growing due to its many potential applications, including rescue operations and pedestrian detection by self-driving cars. However, implementing NLOS imaging on a moving camera remains an open area of research. Existing NLOS imaging methods rely on time-resolved detectors and laser configurations that require precise optical alignment, making it difficult to deploy them in dynamic environments. This work proposes a data-driven approach to NLOS imaging, PathFinder, that can be used with a standard RGB camera mounted on a small, power-constrained mobile robot, such as an aerial drone. Our experimental pipeline is designed to accurately estimate the 2D trajectory of a person who moves in a Manhattan-world environment while remaining hidden from the camera's field-of-view. We introduce a novel approach to process a sequence of dynamic successive frames in a line-of-sight (LOS) video using an attention-based neural network that performs inference in real-time. The method also includes a preprocessing selection metric that analyzes images from a moving camera which contain multiple vertical planar surfaces, such as walls and building facades, and extracts planes that return maximum NLOS information. We validate the approach on in-the-wild scenes using a drone for video capture, thus demonstrating low-cost NLOS imaging in dynamic capture environments.
Differentiable 3D-Gaussian splatting (GS) is emerging as a prominent technique in computer vision and graphics for reconstructing 3D scenes. GS represents a scene as a set of 3D Gaussians with varying opacities and employs a computationally efficient splatting operation along with analytical derivatives to compute the 3D Gaussian parameters given scene images captured from various viewpoints. Unfortunately, capturing surround view ($360^{\circ}$ viewpoint) images is impossible or impractical in many real-world imaging scenarios, including underwater imaging, rooms inside a building, and autonomous navigation. In these restricted baseline imaging scenarios, the GS algorithm suffers from a well-known 'missing cone' problem, which results in poor reconstruction along the depth axis. In this manuscript, we demonstrate that using transient data (from sonars) allows us to address the missing cone problem by sampling high-frequency data along the depth axis. We extend the Gaussian splatting algorithms for two commonly used sonars and propose fusion algorithms that simultaneously utilize RGB camera data and sonar data. Through simulations, emulations, and hardware experiments across various imaging scenarios, we show that the proposed fusion algorithms lead to significantly better novel view synthesis (5 dB improvement in PSNR) and 3D geometry reconstruction (60% lower Chamfer distance).
In this paper, we present a two-stage unsupervised foreground object segmentation network tailored for dynamic scenes affected by atmospheric turbulence. In the first stage, we utilize averaged optical flow from turbulence-distorted image sequences to feed a novel region-growing algorithm, crafting preliminary masks for each moving object in the video. In the second stage, we employ a U-Net architecture with consistency and grouping losses to further refine these masks optimizing their spatio-temporal alignment. Our approach does not require labeled training data and works across varied turbulence strengths for long-range video. Furthermore, we release the first moving object segmentation dataset of turbulence-affected videos, complete with manually annotated ground truth masks. Our method, evaluated on this new dataset, demonstrates superior segmentation accuracy and robustness as compared to current state-of-the-art unsupervised methods.
A good supervised embedding for a specific machine learning task is only sensitive to changes in the label of interest and is invariant to other confounding factors. We leverage the concept of repeatability from measurement theory to describe this property and propose to use the intra-class correlation coefficient (ICC) to evaluate the repeatability of embeddings. We then propose a novel regularizer, the ICC regularizer, as a complementary component for contrastive losses to guide deep neural networks to produce embeddings with higher repeatability. We use simulated data to explain why the ICC regularizer works better on minimizing the intra-class variance than the contrastive loss alone. We implement the ICC regularizer and apply it to three speech tasks: speaker verification, voice style conversion, and a clinical application for detecting dysphonic voice. The experimental results demonstrate that adding an ICC regularizer can improve the repeatability of learned embeddings compared to only using the contrastive loss; further, these embeddings lead to improved performance in these downstream tasks.
Synthetic aperture sonar (SAS) measures a scene from multiple views in order to increase the resolution of reconstructed imagery. Image reconstruction methods for SAS coherently combine measurements to focus acoustic energy onto the scene. However, image formation is typically under-constrained due to a limited number of measurements and bandlimited hardware, which limits the capabilities of existing reconstruction methods. To help meet these challenges, we design an analysis-by-synthesis optimization that leverages recent advances in neural rendering to perform coherent SAS imaging. Our optimization enables us to incorporate physics-based constraints and scene priors into the image formation process. We validate our method on simulation and experimental results captured in both air and water. We demonstrate both quantitatively and qualitatively that our method typically produces superior reconstructions than existing approaches. We share code and data for reproducibility.
Efficient and accurate detection of subtle motion generated from small objects in noisy environments, as needed for vital sign monitoring, is challenging, but can be substantially improved with magnification. We developed a complex Gabor filter-based decomposition method to amplify phases at different spatial wavelength levels to magnify motion and extract 1D motion signals for fundamental frequency estimation. The phase-based complex Gabor filter outputs are processed and then used to train machine learning models that predict respiration and heart rate with greater accuracy. We show that our proposed technique performs better than the conventional temporal FFT-based method in clinical settings, such as sleep laboratories and emergency departments, as well for a variety of human postures.
Mixed reality (MR) is a key technology which promises to change the future of warfare. An MR hybrid of physical outdoor environments and virtual military training will enable engagements with long distance enemies, both real and simulated. To enable this technology, a large-scale 3D model of a physical environment must be maintained based on live sensor observations. 3D reconstruction algorithms should utilize the low cost and pervasiveness of video camera sensors, from both overhead and soldier-level perspectives. Mapping speed and 3D quality can be balanced to enable live MR training in dynamic environments. Given these requirements, we survey several 3D reconstruction algorithms for large-scale mapping for military applications given only live video. We measure 3D reconstruction performance from common structure from motion, visual-SLAM, and photogrammetry techniques. This includes the open source algorithms COLMAP, ORB-SLAM3, and NeRF using Instant-NGP. We utilize the autonomous driving academic benchmark KITTI, which includes both dashboard camera video and lidar produced 3D ground truth. With the KITTI data, our primary contribution is a quantitative evaluation of 3D reconstruction computational speed when considering live video.
Approximately 1.2% of the world's population has impaired voice production. As a result, automatic dysphonic voice detection has attracted considerable academic and clinical interest. However, existing methods for automated voice assessment often fail to generalize outside the training conditions or to other related applications. In this paper, we propose a deep learning framework for generating acoustic feature embeddings sensitive to vocal quality and robust across different corpora. A contrastive loss is combined with a classification loss to train our deep learning model jointly. Data warping methods are used on input voice samples to improve the robustness of our method. Empirical results demonstrate that our method not only achieves high in-corpus and cross-corpus classification accuracy but also generates good embeddings sensitive to voice quality and robust across different corpora. We also compare our results against three baseline methods on clean and three variations of deteriorated in-corpus and cross-corpus datasets and demonstrate that the proposed model consistently outperforms the baseline methods.
Circular Synthetic aperture sonars (CSAS) capture multiple observations of a scene to reconstruct high-resolution images. We can characterize resolution by modeling CSAS imaging as the convolution between a scene's underlying point scattering distribution and a system-dependent point spread function (PSF). The PSF is a function of the transmitted waveform's bandwidth and determines a fixed degree of blurring on reconstructed imagery. In theory, deconvolution overcomes bandwidth limitations by reversing the PSF-induced blur and recovering the scene's scattering distribution. However, deconvolution is an ill-posed inverse problem and sensitive to noise. We propose a self-supervised pipeline (does not require training data) that leverages an implicit neural representation (INR) for deconvolving CSAS images. We highlight the performance of our SAS INR pipeline, which we call SINR, by implementing and comparing to existing deconvolution methods. Additionally, prior SAS deconvolution methods assume a spatially-invariant PSF, which we demonstrate yields subpar performance in practice. We provide theory and methods to account for a spatially-varying CSAS PSF, and demonstrate that doing so enables SINR to achieve superior deconvolution performance on simulated and real acoustic SAS data. We provide code to encourage reproducibility of research.
There is tremendous scope for improving the energy efficiency of embedded vision systems by incorporating programmable region-of-interest (ROI) readout in the image sensor design. In this work, we study how ROI programmability can be leveraged for tracking applications by anticipating where the ROI will be located in future frames and switching pixels off outside of this region. We refer to this process of ROI prediction and corresponding sensor configuration as adaptive subsampling. Our adaptive subsampling algorithms comprise an object detector and an ROI predictor (Kalman filter) which operate in conjunction to optimize the energy efficiency of the vision pipeline with the end task being object tracking. To further facilitate the implementation of our adaptive algorithms in real life, we select a candidate algorithm and map it onto an FPGA. Leveraging Xilinx Vitis AI tools, we designed and accelerated a YOLO object detector-based adaptive subsampling algorithm. In order to further improve the algorithm post-deployment, we evaluated several competing baselines on the OTB100 and LaSOT datasets. We found that coupling the ECO tracker with the Kalman filter has a competitive AUC score of 0.4568 and 0.3471 on the OTB100 and LaSOT datasets respectively. Further, the power efficiency of this algorithm is on par with, and in a couple of instances superior to, the other baselines. The ECO-based algorithm incurs a power consumption of approximately 4 W averaged across both datasets while the YOLO-based approach requires power consumption of approximately 6 W (as per our power consumption model). In terms of accuracy-latency tradeoff, the ECO-based algorithm provides near-real-time performance (19.23 FPS) while managing to attain competitive tracking precision.