Skip to content
View balaboom123's full-sized avatar
11 followers · 5 following

Achievements

Achievement: Pull Sharkx2Achievement: YOLOAchievement: Pair ExtraordinaireAchievement: Starstruckx2

Achievements

Achievement: Pull Sharkx2Achievement: YOLOAchievement: Pair ExtraordinaireAchievement: Starstruckx2

Highlights

  • Pro

Organizations

@NCUE-EE-AIAL

Block or report balaboom123

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
balaboom123/README.md

I'm passionate about Natural Language Processing and Computer Vision, currently pursuing my Master's degree in CS (AI Track).

        

GitHub Statistics

Profile Details GitHub Stats

Research & Publications

ICCE-TW 2026  Towards Compact Sign Language Translation: Frame Rate and Model Size Trade-offs

First author. Studies how input frame rate and model size trade off in gloss-free sign language translation from pose sequences. Reaches competitive results with a 77M-parameter T5-small decoder at reduced frame rates.

ICCE-TW 2026  STRLite: Lightweight Masked Autoencoders for Adaptable Scene Text Recognition

Co-first author. A 6M-parameter ViT-Tiny scene text recognizer trained with MAE pretraining and autoregressive fine-tuning. Reaches 93.8% on the six common STR benchmarks and 81.0% on Union14M.

arXiv 2026   SignDATA: Data Pipeline for Sign Language Translation

First author. A config-driven pipeline that preprocesses 12 sign language datasets into pose landmarks or video clips using MediaPipe, MMPose, and YOLO. Ships WebDataset shards ready for training and has passed 100 stars on GitHub.

ICETA 2024   Two-step Authentication: Multi-biometric System Using Voice and Facial Recognition

First author. A real-time two-step authentication system that combines VGG-16 face recognition with ResCNN voice verification. Published in the IET ICETA 2024 proceedings.

Projects

signdata-slt       tw-exam


SEO-Content-Intelligence-Engine       awesome-italian-brainrot

Pinned Loading

  1. signdata-slt signdata-slt Public

    Modular, config-driven pipeline for preprocessing Sign Language datasets with pose and video outputs using MediaPipe, MMPose, and YOLO..

    Python 101 2

  2. STRLite STRLite Public

    [ICCE-TW 2026] Code base for STRLite: Lightweight Masked Autoencoders for Adaptable Scene Text Recognition

    Python 1

  3. NCUE-EE-AIAL/Two-step-Authentication-Multi-biometric-System NCUE-EE-AIAL/Two-step-Authentication-Multi-biometric-System Public

    [ICETA 2024] A real-time, two-step multi-biometric authentication system combining VGG16-based facial recognition and ResNet-based voice verification.

    Jupyter Notebook 9 1

  4. awesome-italian-brainrot awesome-italian-brainrot Public

    This is a collection to restore all the Italian brainrot.

    Python 265 5

  5. SEO-Content-Intelligence-Engine SEO-Content-Intelligence-Engine Public

    An automated AI data pipeline designed for modern SEO competitor analysis. It leverages Crawl4AI for web scraping, utilizes NLP Entity Extraction & Clustering to reverse-engineer search intent, and…

    TypeScript 1

  6. tw-exam tw-exam Public

    Python