Skip to content
View Jay17205's full-sized avatar

Block or report Jay17205

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.

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
Jay17205/README.md

Nguyễn Tiến Dũng

AI Student at FPT University — Da Nang Campus
Deep Learning · Computer Vision · Time-Series Forecasting · Applied Machine Learning

GitHub Kaggle

About me

  • AI student focused on building reproducible, evidence-backed machine learning projects.
  • Interested in computer vision, deep learning, time-series forecasting, and practical AI deployment.
  • Experienced with experiment design, model comparison, evaluation, error analysis, and technical documentation.
  • Currently building a portfolio for AI/ML internship opportunities.

Technical stack

Python PyTorch TensorFlow OpenCV scikit-learn Pandas NumPy R Docker Git

Featured AI projects

Project What it demonstrates Verified result
Tomato Leaf Disease Classification Fixed-data comparison of lightweight CNNs, strong CNN backbones, and Swin Transformer across six real-world disease classes Swin-Tiny: 87.2% highest single-run accuracy; 84.8% ± 0.8% across three seeds
YOLOv7 Head Detection & People Counting YOLOv7 head detector for CCTV video with per-frame occupancy counting and a reproducible data pipeline 0.981 mAP@0.5, 0.960 precision, 0.949 recall
Hung Yen Water-Level Forecasting Multivariate RNN, LSTM, and Transformer forecasting over 24-, 72-, and 120-hour horizons 24h LSTM: R² 0.9313; best models improve RMSE over persistence by 14.09–19.00%
E-commerce Purchase Intention in R Statistical testing, leakage-safe modeling, threshold tuning, and bootstrap confidence intervals for an imbalanced classification task Decision Tree: F1 0.647, ROC AUC 0.843
CSRNet Crowd Counting in PyTorch Clean CSRNet implementation with density-map generation, training, evaluation, inference, and tests Architecture and pipeline implemented; no unsupported benchmark claim
Tomato Disease AI Web App Streamlit deployment layer for six-class Swin-Tiny inference with validation, Docker, tests, and CI Deployment code linked to the verified tomato benchmark

Core strengths

  • Computer Vision: image classification, object detection, crowd counting, transfer learning, CNNs, Swin Transformer, YOLOv7, CSRNet
  • Time Series: multistep forecasting, RNN, LSTM, Transformer, temporal validation, persistence baselines
  • Machine Learning: preprocessing, feature analysis, class imbalance, statistical testing, model selection, threshold tuning
  • Engineering: reproducible notebooks, testing, CI, GitHub Actions, Docker, Streamlit, clear project documentation

Current direction

I am extending my portfolio from experimental notebooks into reproducible AI repositories and deployable demos. My next goals include tracking-based people counting with ByteTrack, model explainability, and lightweight deployment.


Open to AI/ML internship opportunities and technical collaboration.

Pinned Loading

  1. tomato-leaf-disease-classification tomato-leaf-disease-classification Public

    Reproducible deep learning benchmark for six-class tomato leaf disease classification under real-world conditions.

    Jupyter Notebook

  2. yolov7-head-detection-counting yolov7-head-detection-counting Public

    YOLOv7 head detection and per-frame people counting for indoor CCTV footage.

    Python

  3. ecommerce-purchase-intention-r ecommerce-purchase-intention-r Public

    Reproducible R project for e-commerce purchase intention analysis and conversion prediction.

    R