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

Modern deep learning models are inefficient: they are trained on many common-scene images with large uninformative regions.

My research addresses this inefficiency from two angles:

• Data efficiency: synthesizing or collecting long-tail scenes (e.g., TPSeNCE, ROADWork)
• Compute efficiency: oversampling salient regions (e.g., Instance-Warp, WarpI2I), and reducing spatial redundancy (e.g., MapThunder)

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

We build an online vectorized HD map construction method with fewer tokens, faster inference, and better accuracy.

We warp the input image to enlarge salient regions (e.g., faces, eyes, objects) to better preserve fine details in the compressed latent space during I2I.

We introduce ROADWork, a large-scale open-source dataset and benchmark with fine-grained annotations and scene descriptions for driving in work zones.

We warp the image at instance-level to oversample foreground objects and undersample background regions to improve domain adaptation.

We introduce TPSeNCE, a rain generation framework with Triangular Probability Similarity (TPS) and Semantic Noise Contrastive Estimation (SeNCE) for realistic rainy scene synthesis.

Present a comprehensive survey of low-light image enhancement (LLIE) and introduce Night Wenzhou, a large-scale, high-resolution video dataset captured in fast motion with diverse illuminations and degradation.

PointNorm, a point cloud analysis network with a DualNorm module that leverages local mean and global standard deviation to address irregular (i.e., unevenly distributed) point clouds.

We introduce SGZ, a zero-shot low-light image and video enhancement framework at 1000 FPS with pixel-wise light deficiency estimation, recurrent image enhancement, and unsupervised semantic segmentation

We 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.

Mentor: Guohao Zhang

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

Director: Dr. Feng Guo

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

Director: Dr. Wangjiang Zhu

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