Xiangbo Gao-image
about-me-image

Xiangbo Gao

I love to use deep learning to solve real-world problems.

about-me-image

About me

I am a current masterstudent at the University of Micigan - Ann Arbor. My research focus on deep learning and its applications such as computer vision and autonomous system. I am currently looking for internship opportunities in Robotics and Autonomous Driving.

  • Location:Ann Arbor, MI
  • Age:23
  • Nationality:China
  • Interests:Skiing, Rock climbing
  • Study:University of Micigan - Ann Arbor

Resume

Education

Robotics M.S. Student

University of Michigan, Ann Arbor2023.9 - PRESENT

B.S. in Computer Science | B.S. in Mathematics

University of California, Irvine2018.9 - 2023.3

GPA 3.72/4

Summer Session

University of California, Berkeley2019.6 - 2019.9

GPA 3.566/4

Research

Multi-modal 3D Object Detection in autonomous driving scenario

Independent2022.6 - Present
  • -iterature review of existing general 3D object detection algorithms including camera-based, point-based, voxel-based, and multi-model algorithms.
  • -Reproduce and improve existing camera-only and multi-model 3D object detection algorithms (Pseudo-LiDAR, Lidar Painting, Bev fusion).

Auto-generated graphical model in the autonomous driving system

Advanced Integrated Cyber-Physical Systems Lab, University of California, Irvine2022.5 - Present
  • -Designed and created a multi-domain autonomous driving dataset for different driving scenarios using CARLA simulator.
  • -Designed a probabilistic LSTM structure that encodes states to variational embeddings. Transferred the knowledge from PointNet to encode the unordered lane marks information.
  • -Evaluating the model performance and cross-domain transferability by various metrics and comparing them with other motion prediction algorithms (MTP, PGP, Trajectron++, etc.)

Goal-conditional Reinforcement Learning

Intelligent Dynamics Lab, University of California, Irvine2022.2 - 2022.6
  • -Reviewed literature on imitation learning and general reinforcement learning.
  • -Come up with spring loss which uses the idea of contrastive learning that aligns the embeddings in linear distance.
  • -Visualization by the PCA dimension reduction method shows that the learned embedding better aligns with the real-world trajectory.

Adversarial Attack with Semantic Pattern

Institute of Computer Vision, Shenzen University, China2021.4 - 2022.11

Long-tailed Cervical Cell Detection

Institute of Computer Vision, Shenzen University, China2021.4 - 2022.1

ZerO Waste Anteaters

Donald Bren Hall, University of California, Irvine2020.9 - 2022.4
  • -Led a team of 8 members to explore waste recognition solutions.
  • -Designed a computer vision and waste recognition tutorial that stimulate students' interests.
  • -Trained light-weight models for waste image classification (Mobilenetv3, ShuffleNet, and EfficientNet) and waste object detection (Faster-RCNN and YOLOv5); achieved ~0.94 classification accuracy and ~0.76 mean precision error.
  • -Deployed the waste recognition models to resource-limited machines (Jetson Nano)

Employment

Perception Research Intern

Anhui Cowa ROBOT Co., Ltd, Shanghai, China2023.4 - 2023.7
  • -Online HD Map Construction with Flow Map Prior (3.2% higher mAP than the baseline)
    • ---Propose to use historical vehicle trajectories as prior
    • ---Reproduce the neural map prior and allivate the catastrophic forgetting problem by adding noise and dummy features.
    • ---Propose Keypoints DTW Loss to increase the consistency of the regression loss
    • ---Inspired by BevFormerV2, implement 2D auxiliary keypoints detection to further boost the accuracy.
  • -Motion Prediction with Historical Trajectories Clustering

Full-stack Software developer

Tandll Investment Management Limited, China2020.6 - 2020.8

-Built a quantitative trading support website using Python (Django & React) and MySQL, which supported high-level trading management, model parameters modification, and historical data & behaviors Visualization

VR Software developer

Calit 2, University of California, Irvine2019.2 - 2019.7

-Worked with a team of 6 to Develop a simple-to-understand VR teaching aid of MA6 Mask Aligner, which enables students to use the machine without physically getting into the clean room.

Competitions

CVPR Camera-based online HD map construction challenge 2023

N/A2023.5

Result: Rank 13th in CVPR Camera-based online HD map construction challenge 2023

UCI 2020 Machine Learning Hackathon

University of California, Irvine, CA, USA2020.4

1st place on the subproject of 3D Human Pose with Scene Constraints

Google Hash Code 2020 Algorithms Competition

Irvine, CA2020.2

Result: 2nd place / 13 at UCI | Team name: ε=.99

Netease Hackathon Competition

China2020.6

Outstanding Award

Publications

Selected

Sample Hardness Based Gradient Loss for Long-Tailed Cervical Cell Detection

Minmin Liu, Xuechen Li, Xiangbo Gao, Junliang Chen, Linlin Shen, Huisi Wu

Medical Image Computing and Computer Assisted Intervention – MICCAI 2022

Due to the difficulty of cancer samples collection and annotation, cervical cancer datasets usually exhibit a long-tailed data distribution. When training a detector to detect the cancer cells in a WSI (Whole Slice Image) image captured from the TCT (Thinprep Cytology Test) specimen, head categories (e.g. normal cells and inflammatory cells) typically have a much larger number of samples than tail categories (e.g. cancer cells). Most existing state-of-the-art long-tailed learning methods in object detection focus on category distribution statistics to solve the problem in the long-tailed scenario without considering the "hardness" of each sample. To address this problem, in this work we propose a Grad-Libra Loss that leverages the gradients to dynamically calibrate the degree of hardness of each sample for different categories, and re-balance the gradients of positive and negative samples. Our loss can thus help the detector to put more emphasis on those hard samples in both head and tail categories. Extensive experiments on a long-tailed TCT WSI image dataset show that the mainstream detectors, e.g. RepPoints, FCOS, ATSS, YOLOF, etc. trained using our proposed Gradient-Libra Loss, achieved much higher (7.8%) mAP than that trained using cross-entropy classification loss.

about-me-image

On Submission

Scale-free and Task-agnostic Attack: Generating Photo-realistic Adversarial Patterns with Patch Quilting Generator

Xiangbo Gao, Cheng Luo, Qinliang Lin, Weicheng Xie, Minmin Liu, Linlin Shen, Keerthy Kusumam, Siyang Song

arXiv:2208.06222

Traditional L_p norm-restricted image attack algorithms suffer from poor transferability to black box scenarios and poor robustness to defense algorithms. Recent CNN generator-based attack approaches can synthesize unrestricted and semantically meaningful entities to the image, which is shown to be transferable and robust. However, such methods attack images by either synthesizing local adversarial entities, which are only suitable for attacking specific contents or performing global attacks, which are only applicable to a specific image scale. In this paper, we propose a novel Patch Quilting Generative Adversarial Networks (PQ-GAN) to learn the first scale-free CNN generator that can be applied to attack images with arbitrary scales for various computer vision tasks. The principal investigation on transferability of the generated adversarial examples, robustness to defense frameworks, and visual quality assessment show that the proposed PQG-based attack framework outperforms the other nine state-of-the-art adversarial attack approaches when attacking the neural networks trained on two standard evaluation datasets (i.e., ImageNet and CityScapes).

about-me-image

Get in touch.