Ardian Umam

Short bio

My name is Ardian Umam (禹安銳), you can call me Ardian, or 安銳 (An-Rui) in Chinese. I am currently pursuing my Ph.D. in the Department of Electrical Engineering and Computer Science at National Yang Ming Chiao Tung University, Taiwan, working with Prof. Yen-Yu Lin from VLLab and Prof. Jen-Hui Chuang from Islab.

My reseach interests lie in (but not limited to) deep learning, computer vision, natural language processing, and multi-modal AI. Over the past 8 years, I have been working on various topics which involve 1D data (audio and languange), 2D data (image) and 3D data (point cloud and mesh). The tasks include audio quality estimation, optical character recognition, camera calibration, 3D point cloud augmentation, 3D segmentation, unsupervised/weakly supervised segmentation, multi-modal (vision-language) recognition, and large language models using RAG (Retrieval Augmented Generation). Additionally, I interned at Google with the ChromeOS Audio Team and at ITRI working on model compression.

Selected publications

PartDistill: 3D Shape Part Segmentation by Vision-Language Model Distillation

Ardian Umam, Cheng-Kun Yang, Min-Hung Chen, Jen-Hui Chuang, and Yen-Yu Lin

In IEEE/CVF International Conference on Computer Vision (CVPR) , 2024

arXiv Bib Code HTML Project Page

@inproceedings{umam2023partdistill,
  title = {PartDistill: 3D Shape Part Segmentation by Vision-Language Model Distillation},
  author = {Umam, Ardian and Yang, Cheng-Kun and Chen, Min-Hung and Chuang, Jen-Hui and Lin, Yen-Yu},
  booktitle = {IEEE/CVF International Conference on Computer Vision (CVPR)},
  year = {2024},
}

Unsupervised Point Cloud Co-part Segmentation via Co-attended Superpoint Generation and Aggregation

Ardian Umam, Cheng-Kun Yang, Jen-Hui Chuang, and Yen-Yu Lin

IEEE Transactions on Multimedia (TMM), 2024

Bib HTML

@article{umam2024unsupervised,
  title = {Unsupervised Point Cloud Co-part Segmentation via Co-attended Superpoint Generation and Aggregation},
  author = {Umam, Ardian and Yang, Cheng-Kun and Chuang, Jen-Hui and Lin, Yen-Yu},
  journal = {IEEE Transactions on Multimedia (TMM)},
  year = {2024},
  publisher = {IEEE},
}

Point MixSwap: Attentional Point Cloud Mixing via Swapping Matched Structural Divisions

Ardian Umam, Cheng-Kun Yang, Yung-Yu Chuang, Jen-Hui Chuang, and Yen-Yu Lin

In European Conference on Computer Vision (ECCV) , 2022

Bib Code HTML

@inproceedings{umam2022point,
  title = {Point MixSwap: Attentional Point Cloud Mixing via Swapping Matched Structural Divisions},
  author = {Umam, Ardian and Yang, Cheng-Kun and Chuang, Yung-Yu and Chuang, Jen-Hui and Lin, Yen-Yu},
  booktitle = {European Conference on Computer Vision (ECCV)},
  pages = {596--611},
  year = {2022},
  organization = {Springer},
}

Geometry-based Camera Calibration Using Closed-form Solution of Principal Line

Jen-Hui Chuang, Chih-Hui Ho, Ardian Umam, Hsin-Yi Chen, Jenq-Neng Hwang, and Tai-An Chen

IEEE Transactions on Image Processing (TIP), 2021

arXiv Bib HTML

@article{chuang2021geometry,
  title = {Geometry-based Camera Calibration Using Closed-form Solution of Principal Line},
  author = {Chuang, Jen-Hui and Ho, Chih-Hui and Umam, Ardian and Chen, Hsin-Yi and Hwang, Jenq-Neng and Chen, Tai-An},
  journal = {IEEE Transactions on Image Processing (TIP)},
  volume = {30},
  pages = {2599--2610},
  year = {2021},
  publisher = {IEEE},
}

Education

National Yang Ming Chiao Tung University
PhD in Computer Science, 2020 - Now
Join Vision and Learning Lab. Thesis: 3D recognition under low annotation costs
National Chiao Tung University
MSc in Computer Science
Join Intelligent System Lab. Thesis: A light deep learning based method for bank serial number recognition
Gadjah Mada University
BSc in Electrical Engineering
Thesis: Adaptive-PID control system for dc motor speed

Work experience

Lecturer
Institut Teknologi Bandung, 2019 - Now
Faculty member in School of Electrical Engineering and Informatics
Graduate Intern
Google, Apr - Dec 2022
Reduced audio quality estimation error for ChromeOS by 45% using a self-supervised learning approach. The method leverages large-scale unlabelled audio data from publicly available datasets and internal company datasets to improve the feature representations (audio encoder)
AI Engineer
Computer Vision Research Center - NCTU, 2018 - 2019
Optimized deep learning models for depth map estimation (20 to 40 FPS) and object detection (5 to 20 FPS) on edge device (Jetson TX2) through architectural downscaling and TensorRT optimization
Summer Intern
ITRI (Industrial Technology Research Institute), Jul - Aug 2018
Studied deep learning computational reduction techniques, e.g., network pruning, from recent papers
Avionic Engineer
LAPAN (National Institute of Aeronautics and Space), 2015 - 2016
Developed a UAV telemetric monitoring system to track power consumption and UAV states

Award

CTCI Research Award 2024

Awarded by The CTCI Foundation to recognize outstanding PhD students in Taiwan

Doctoral Scholarship Award

Awarded a doctoral (full) scholarship by the Ministry of Education - Taiwan

Best M.S. Student Award

Awarded as the best master student of the Electrical Engineering and Computer Science Department in the commencement day of NCTU, 2018

TOP1 Final Project Competition

Line Following Robot competition, by "Deep Learning Course (ECM9042)", NCTU Fall Semester, 2020

TOP1 in-class Kaggle Competition

Best entry in the leader board of in-class kaggle competition about movie recommendation system held by “Cloud Computing and Big Data Analytics Class” – NCTU, Spring 2017

Siswa Teladan Putra 1 - Klaten

Annual event held by Dinas Pendidikan Klaten to select student representing Siswa Teladan in province level, Central Java

Community service

Conference reviewer

CVPR, ICCV, ECCV, WACV, AAAI, IJCAI, ICRA

Journal reviewer

TPAMI, IJCV, CVIU, ACM CSUR