Jiayi Wu

Jiayi Wu

Auto-generated profile, based on publicly available information

Jiayi Wu

University of Maryland, USA

Research Focus & Metrics

Academic Career Length

4Years (2023-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
29 May 2023
Verified
Conference Paper
Open access
BYBoxiao Yu
Boxiao Yu
JWJiayi Wu
Jiayi Wu
JIJahidul Islam
Jahidul Islam
In this paper, we present a fast monocular depth estimation method for enabling 3D perception capabilities of low-cost underwater robots. We formulate a novel end-to-end deep visual learning pipeline named UDepth, which incorporates domain knowledge of image formation characteristics of natural underwater scenes. First, we adapt a new input space from raw RGB image space by exploiting underwater light attenuation prior, and then devise a least-squared formulation for coarse pixel-wise depth prediction. Subsequently, we extend this into a domain projection loss that guides the end-to-end learning of UDepth on over 9K RGB-D training samples. UDepth is designed with a computationally light MobileNetV2 backbone and a Transformer-based optimizer for ensuring fast inference rates on embedded systems. By domain-aware design choices and through comprehensive experimental analyses, we demonstrate that it is possible to achieve state-of-the-art depth estimation performance while ensuring a small computational footprint. Specifically, with 70 % −80 % less network parameters than existing benchmarks, UDepth achieves comparable and often better depth estimation performance. While the full model offers over 66 FPS (13 FPS) inference rates on a single GPU (CPU core), our domain projection for coarse depth prediction runs at 51.5 FPS rates on single-board Jetson TX2s. The inference pipelines are available at https://github.com/uf-robopi/UDepth.

1 January 2025 in IEEE Transactions on Pattern Analysis and Machine Intelligence
Verified
Journal Article
Closed access
JWJiayi Wu
Jiayi Wu
TWTianfu Wang
Tianfu Wang
MSMd. Abu Bakr Siddique
Md. Abu Bakr Siddique
JIJahidul Islam
Jahidul Islam
CFCornelia Fermuller
Cornelia Fermuller
CMChris Metzler
Chris Metzler
Underwater image restoration algorithms seek to restore the color, contrast, and appearance of a scene that is imaged underwater. They are a critical tool in applications ranging from marine ecology and aquaculture to underwater construction and archaeology. While existing pixel-domain diffusion-based image restoration approaches are effective at restoring simple scenes with limited depth variation, they are computationally intensive and often generate unrealistic artifacts when applied to scenes with complex geometry and significant depth variation. In this work we overcome these limitations by combining a novel network architecture (SLURPP) with an accurate synthetic data generation pipeline. SLURPP combines pretrained latent diffusion models-which encode strong priors on the geometry and depth of scenes-with an explicit scene decomposition-which allows one to model and account for the effects of light attenuation and backscattering. To train SLURPP we design a physics-based underwater image synthesis pipeline that applies varied and realistic underwater degradation effects to existing terrestrial image datasets. This approach enables the generation of diverse training data with dense medium/degradation annotations. We evaluate our method extensively on both synthetic and real-world benchmarks and demonstrate state-of-the-art performance. Notably, SLURPP is over $200\times$ faster than existing diffusion-based methods while offering $\sim 3 dB$ improvement in PSNR on synthetic benchmarks. It also offers compelling qualitative improvements on real-world data. Project website https://tianfwang.github.io/slurpp/.

12 March 2025 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
ASAbu Bakr Siddique
Abu Bakr Siddique
JWJiayi Wu
Jiayi Wu
IRIoannis Rekleitis
Ioannis Rekleitis
JIJahidul Islam
Jahidul Islam
In this letter, we introduce the idea of AquaFuse, a physics-based method for synthesizing waterbody properties in underwater imagery. We formulate a closed-form solution for waterbody fusion that facilitates realistic data augmentation and geometrically consistent underwater scene rendering. AquaFuse leverages the physical characteristics of light propagation underwater to synthesize the waterbody from one scene to the object contents of another. Unlike data-driven style transfer methods, AquaFuse preserves the depth consistency and object geometry in an input scene. We validate this unique feature by comprehensive experiments over diverse sets of underwater scenes. We find that the AquaFused images preserve over 94% depth consistency and 90–95% structural similarity of the input scenes. We also demonstrate that it generates accurate 3D view synthesis by preserving object geometry while adapting to the inherent waterbody fusion process. AquaFuse opens up a new research direction in data augmentation by geometry-preserving style transfer for underwater imaging and robot vision.

7 December 2023 in IEEE Transactions on Geoscience and Remote Sensing
Verified
Journal Article
Closed access
ARA. Kaleo Roberts
A. Kaleo Roberts
JWJiayi Wu
Jiayi Wu
AMAlejandro Monsivais-Huertero
Alejandro Monsivais-Huertero
JJJasmeet Judge
Jasmeet Judge
RMRobert C. Moore
Robert C. Moore
KSKamal Sarabandi
Kamal Sarabandi
Satellite and airborne radars currently monitor agricultural regions on Earth. Corn is a globally important crop that may benefit from radar observations for estimating soil moisture (SM) and other quantities. Estimates of SM could be used to enhance crop yield and aid in weather prediction. A scattering model is needed, however, to accurately estimate these quantities. Historically, corn is a difficult crop to model at microwave frequencies, and only approximate models for it exist. Novel models based on full-wave electromagnetic solvers can be more accurate by accounting for multiple scattering among plant constituents, other adjacent plants, and the underlying soil surface. Such a model is computationally expensive, but the increased availability of computing resources may make it more feasible. This article presents a model for corn at L-band based on finite element method (FEM) simulations in conjunction with Monte Carlo methods to estimate polarimetric backscattering coefficients. The FEM simulation uses periodic boundary conditions to limit its size. The physical representation of the corn plants comes from data-based 3-D plant models. The results of simulations are validated with synthetic aperture radar (SAR) data obtained during the SM active passive validation experiment of 2012 (SMAPVEX12) experimental campaign. The estimated backscattering coefficients of the SAR data are within ±2 dB for all polarization channels. Validation is performed for two days within the experimental campaign. Good agreement is observed between the simulated and measured values. This result indicates that the model can give novel insights into the scattering characteristics of corn. Future work remains to build an invertible model for estimating SM from backscatter measurements.

10 July 2024 in IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing
Verified
Journal Article
Open access
KNKaiser Niknam
Kaiser Niknam
JJJasmeet Judge
Jasmeet Judge
ARA. Kaleo Roberts
A. Kaleo Roberts
AMAlejandro Monsivais-Huertero
Alejandro Monsivais-Huertero
RMRobert C. Moore
Robert C. Moore
JWJiayi Wu
Jiayi Wu
Microwave remote sensing offers a powerful tool for monitoring the growth of short, dense vegetation such as soybeans. As the plants mature, changes in their biomass and 3-D structure impact the electromagnetic (EM) backscatter signal. This backscatter information holds valuable insights into crop health and yield, prompting the need for a comprehensive understanding of how structural and biophysical properties of soybeans as well as soil characteristics contribute to the overall backscatter signature. In this study, a full-wave model is developed for simulating L-band backscatter from soybean fields. Leveraging the ANSYS high-frequency structure simulator (HFSS) framework, the model solves for the scattering of EM waves from realistic 3-D structural models of soybean, explicitly incorporating the interplant scattering effects. The model estimates of backscatter match well with the field observations from the SMAPVEX16-MicroWEX and SMAPVEX12, with average differences of 1–2 dB for co-pol and less than 4 dB for cross-pol. Furthermore, the model effectively replicates the temporal dynamics of crop backscatter throughout the growing season. The HFSS analysis revealed that the stems, pods, and soil are the primary contributors to HH-pol backscatter, while the branches contribute to VV-pol, and leaves impact the cross-pol signatures. In addition, a sensitivity study with a 3-D bare soil surface resulted in an average variation of 8 dB in co- and cross-pol, even when the root-mean-square height and correlation length were held constant. These capabilities underscore the model's potential to provide insights into the underlying dynamics of the backscatter for growing vegetation.

20 April 2026 in IEEE Robotics and Automation Letters
Verified
Journal Article
Closed access
NRNaitri Rajyaguru
Naitri Rajyaguru
TWTianfu Wang
Tianfu Wang
ATAryan Tajne
Aryan Tajne
BHBotao He
Botao He
JWJiayi Wu
Jiayi Wu
Yiannis AloimonosYiannis Aloimonos
Yiannis Aloimonos
Glass surfaces remain challenging for indoor robot perception. Depth sensors and RGB-only monocular depth estimation often fail because of reflections, refractions, and low-texture regions. To this end, we present PolarDepth, a polarization-enhanced monocular depth framework for glass-dominant environments. We utilize a single polarization sensor to obtain a standard RGB image and a three-channel encoding of polarization cues, designed to be compatible with RGB-trained foundation depth models, and which we call Polar-RGB. This representation enables recovery of structure in transparent and specular regions where RGB cues are unreliable. We predict depth from both the RGB and Polar-RGB representations and fuse the predictions using a learned per-pixel reliability gate for mid-level fusion, highlighting RGB in diffuse regions and polarization on reflective surfaces. We demonstrate improvements in depth estimation and visual odometry performance over RGB-only baselines in glass-walled corridors on both real-world and synthetic data.