Shen Zheng 「郑深」

I am a second-year PhD student at the Robotics Institute, Carnegie Mellon University (CMU), advised by Dr. Srinivasa Narasimhan. I have interned at Waymo and Momenta, and worked at Lucid Motors. I completed my M.S. in Computer Vision (MSCV) at CMU, and earned my B.S. in Mathematics from Wenzhou-Kean University (WKU), where I worked with Dr. Gaurav Gupta.

Email1: shenzhen@andrew.cmu.edu

Email2: lebronshenzheng@gmail.com

Resume  /  Google Scholar  /  Github  /  Leetcode  /  YouTube

profile photo
Research Areas

My current research takes a data-centric approach to enable robust understanding of long-tail yet safety-critical scenarios in autonomous driving (e.g., low-light, bad weather, work zones).

  • Collected Data (采集数据): ROADWork
  • Synthesized Data (生成数据): TPSeNCE
  • Augmented Data (增强数据): Instance-Warp
  • Historical Data (历史数据): work in progress
  • Predicted Data (预测数据): future work

My earlier works focused on image restoration and enhancement (e.g., SGZ, LLIE_Survey).

Selected Publications

ROADWork Dataset: Learning to Recognize, Observe, Analyze and Drive Through Work Zones
Anurag Ghosh, Shen Zheng, Robert Tamburo, Juan R. Alvarez Padilla, Hailiang Zhu, Michael Cardei, Nicholas Dunn, Christoph Mertz, Srinivasa Narasimhan
ICCV 2025
[Paper] [Webpage] [GitHub]

Motivation: Navigating through work zones is challenging due to a lack of large-scale open datasets.

Solution: Introduce the ROADWork dataset, which is so far the largest open-source work zone dataset, to help learn how to recognize, observe, analyze, and drive through work zones.

Instance-Warp: Saliency Guided Image Warping for Unsupervised Domain Adaptation
Shen Zheng★, Anurag Ghosh★, Srinivasa Narasimhan
WACV 2025
[Paper] [Webpage] [Code]

Motivation: Domain adaptation methods struggle to learn smaller objects amidst dominant backgrounds with high cross-domain variations.

Solution: Warp source-domain images in-place using instance-level saliency to oversample objects and undersample backgrounds during domain adaptation training.

TPSeNCE: Towards Artifact-Free Realistic Rain Generation for Deraining and Object Detection in Rain
Shen Zheng, Changjie Lu, Srinivasa Narasimhan
WACV 2024
[Paper] [Webpage] [Code] [Slides] [Poster]

Motivation: Previous image-to-image translation methods produce artifacts and distortions, and lack control over the amount of rain generated.

Solution: Introduce a Triangular Probability Similarity (TPS) loss to minimize the artifacts and distortions during rain generation. Propose a Semantic Noise Contrastive Estimation (SeNCE) strategy to optimize the amounts of generated rain. Show that realistic rain generation benefits deraining and object detection in rain.

Low-Light Image Enhancement: A Comprehensive Survey and Beyond
Shen Zheng, Yiling Ma, Jinqian Pan, Changjie Lu, Gaurav Gupta

[Paper] [Code]

Motivation: Existing LLIE datasets focus on either overexposure or underexposure, not both, and usually feature minimally degraded images captured from static positions.

Solution: Present a comprehensive survey of low-light image enhancement (LLIE). Propose the SICE_Grad and SICE_Mix image datasets, which include images with both overexposure and underexposure. Introduce Night Wenzhou, a large-scale, high-resolution video dataset captured in fast motion with diverse illuminations and degradation.

PointNorm: Dual Normalization is All You Need for Point Cloud Analysis
Shen Zheng, Jinqian Pan, Changjie Lu, Gaurav Gupta
IJCNN 2023 (Oral Presentation)
[Paper] [Webpage] [Code] [Slides]

Motivation: Current point cloud analysis methods struggle with irregular (i.e., unevenly distributed) point clouds.

Solution: PointNorm, a point cloud analysis network with a DualNorm module (Point Normalization & Reverse Point Normalization) that leverages local mean and global standard deviation.

Semantic-Guided Zero-Shot Learning for Low-Light Image/Video Enhancement
Shen Zheng, Gaurav Gupta
WACV 2022
[Paper] [Webpage] [Code] [Slides]

Motivation: Current low-light image enhancement methods cannot handle uneven illuminations, is computationally inefficient, and fail to preserve the semantic information.

Solution: Introduce SGZ, a zero-shot low-light image enhancement framework with pixel-wise light deficiency estimation, parameter-free recurrent image enhancement, and unsupervised semantic segmentation.

AS-IntroVAE: Adversarial Similarity Distance Makes Robust IntroVAE
Changjie Lu, Shen Zheng, Zirui Wang, Omar Dib, Gaurav Gupta
ACML 2022
[Paper] [Code] [Slides]

Motivation: Generative models experience posterior collapse and vanishing gradient due to no effective metric for real-fake image evaluation.

Solution: Propose Adversarial Similarity Distance Introspective Variational Autoencoder (AS-IntroVAE), which can address the posterior collapse and the vanishing gradient problem in image generation in one go.

Professional Experiences
Perception Software Engineer (Intern) at Waymo

Mentor: Guohao Zhang

(WIP) Improved Online HD Map Construction using long-term and short-term memory fusion.

Perception Software Engineer (Full-Time) at Lucid Motors

Director: Dr. Feng Guo

Working as a perception software engineer in the ADAS perception team responsible for auto-parking, traffic light detection, and blockage detection.

Improved BEVFormer for auto-parking (reverse & parallel) by using extrinsic calibration to interpolate and smooth edges to enhance curb detection.

Trained YOLO6 on full-resolution images containing traffic lights and fine-tuned arrow types, confidence, IoU, and area thresholds, resulting in a 40+% improvement in mAP (final mAP: 98%+ for day; 90%+ for night).

Developed a binary semantic segmentation model based on CenterNet to detect blockages such as ice, snow, mud, mud blur, rain drops, and sun glares, achieving 93%+ IoU.

Perception Engineer (Intern) at Momenta

Director: Dr. Wangjiang Zhu

Responsible for long-tailed data augmentation, training data auto-labeling and cleaning, and model evaluation for traffic light detection algorithms.

Implemented CycleGAN to conduct unsupervised data augmentation, converting traffic light bulbs from left arrow to round & leftUturn arrow.

Constructed a traffic light auto-label model using quantized VoVNet-57, filtering 14,618 incorrect annotations from 1,160,513 labeled frames.

Increased the classification accuracy for leftUturn traffic light from 78.41% to 87.27%, and the mean average precision from 93.01% to 94.80%.

Services
Technical Program Committee:
WCCI 2024

Conference Reviewers:
CVIP (2021, 2022), AAAI (2022), IJCNN (2023, 2024, 2025), WACV (2023, 2024, 2025), ECCV (2024), CVPR (2025,2026), ICCV (2025)

Journal Reviewers:
TNNLS, IJCV, TCSVT, ESWA, EAAI, JVCIR, Neurocomputing
Co-Instructor at Wenzhou-Kean University

Course: MATH 3291/3292 (Computer Vision)

Slide | Recordings
Invited Speaker at Fudan University

Topic: Image Processing with Machine Learning
Content Creator:
Made 100+ YouTube video solutions for Leetcode algorithms questions.
Skills
Programming Languages:
Python, R, Java, C++, Matlab, HTML, Mathematica, Shell, LaTeX, Markdown

Frameworks & Platforms:
Pytorch, TensorFlow, Keras, Ubuntu, Docker, Git, ONNX, CUDA

Libraries:
Scikit-Learn, SciPy, NumPy, OpenCV, Matplotlib, Pandas

Fun Facts
Languages:
Chinese, English

Sports:
Basketball, Table Tennis, Swimming, Cycling, Hiking, Weightlifting

Games:
DOTA2, AOE2, Warcraft III

Beliefs: