Skip to content

forallx94/Portfolio

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

13 Commits
 
 
 
 
 
 

Repository files navigation

English | 한국어

AI Researcher Portfolio

An AI researcher with extensive experience in AI image processing and data analysis. I have successfully executed projects across various domains, including computer vision, natural language processing, and time-series data analysis.


Work Experience

AITHENUTRIGENE

AI Researcher (April 2023 - Present)

Key Responsibilities

  • As a member of the AI Image Team, I was responsible for developing various computer vision models for classification, object detection, and OCR.

Projects

1. Azure OCR

Period: April 2023 - July 2023

Azure OCR1 Azure OCR2

Project Details
  • Objective: To accurately extract information from in-store product price tags by capturing photos.
  • Key Achievements:
    • Collected image data of price tags within stores.
    • Detected the location of price tags in images using Azure Computer Vision (Object Detection).
    • Extracted text from within the price tags using Azure OCR.

2. Shelf Object Detection

Period: July 2023 - November 2023

shelf_object_detection

Project Details
  • Objective: To detect and classify products on snack shelves in stores.
  • Key Achievements:
    • Collected image data of snack shelves in stores.
    • Detected snack objects on the shelves using the DINO model (detrex library).
    • Classified the detected objects to identify product names using the ViT model (timm library).

3. Commercial Vehicle Classification

Period: November 2023 - August 2024

Commercial vehicle

Project Details
  • Objective: To verify the consistency between photos of commercial vehicles and their entered information using AI.
  • Key Achievements:
    • Collected 4 vehicle images (front, back, left, right) and 3 pieces of vehicle information (type, license plate, VIN) via an app.
    • Performed vehicle classification by comprehensively analyzing the four images using the Swin-Transformer V2 model.
    • Applied a semi-supervised learning-based anomaly detection model.
    • Recognized text from license plates and Vehicle Identification Numbers (VIN) using PaddleOCR.

4. Power Consumption Prediction for Automotive Parts Forging Lines

Period: November 2023 - December 2025

Power Prediction

Project Details
  • Objective: To develop a solution for predicting and reducing power consumption to improve energy efficiency in automotive parts forging lines.
  • Key Achievements:
    • Collected and analyzed time-series data (power, temperature, etc.) from the forging lines.
    • Predicted future power consumption using Time-Series Forecasting models.
    • Identified energy-inefficient sections and proposed reduction measures based on the prediction model.

SPILAB

AI Researcher (July 2020 - November 2021)

Key Responsibilities

  • As a founding member of the startup, I collaborated with the CEO on project planning and coordination of external projects.
  • Led and executed core development tasks.

Projects

1. Car Plate OCR

Project Details
  • Objective: To automatically recognize the type and license plate of vehicles entering electric vehicle charging stations.
  • Key Achievements:
    • Collected image data using a Raspberry Pi 4.
    • Generated synthetic electric vehicle license plate data using Image Augmentation techniques.
    • Developed a vehicle license plate detection model based on YOLO v4 and v5.
    • Developed an OCR model based on Bidirectional LSTM.
  • Link: https://github.com/forallx94/Electronic-Car-Generate

2. Grass Disease Detection

grass_disease_detection

Project Details
  • Objective: To detect the condition of grass and the presence of diseases by analyzing drone-captured images of a golf course.
  • Key Achievements:
    • Extracted image frames from drone-recorded videos.
    • Detected grass diseases using the EfficientNet B5 model.
    • Managed and stored the analysis results in MongoDB.

3. Cancer Clinical Trials Eligibility

CCTE

Project Details
  • Objective: To provide information on suitable clinical trials for new patients by leveraging cancer clinical trial data.
  • Key Achievements:
    • Preprocessed and structured clinical trial data.
    • Performed text embedding using TF-IDF, Word2Vec, and BERT.
    • Conducted network analysis and clustering using HDBSCAN and Hierarchical Clustering.
    • Developed a core keyword extraction algorithm.

4. Smart Factory IoT Unsupervised Anomaly Detection

samrt_facory

Project Details
  • Objective: To develop an unsupervised learning model that predicts anomalies in advance using IoT data from factory equipment.
  • Key Achievements:
    • Retrieved data from factory compressors and rectifiers from MySQL.
    • Performed time-series data preprocessing and feature engineering.
    • Developed an unsupervised anomaly detection model using Machine Learning, Prophet, and Autoencoder.
  • Link: https://github.com/forallx94/Sequential_Anomaly_detecion

5. BAMS/HAMS Energy Consumption Forecast

Project Details
  • Objective: To predict energy consumption in homes and buildings, including temperature, humidity, electricity usage, and hot water usage.
  • Key Achievements:
    • Preprocessed energy consumption data for houses and buildings.
    • Developed an energy consumption forecasting model using Prophet and Residual LSTM.
    • Uploaded the prediction results to MongoDB.

HERSS

AI Researcher (March 2020 - June 2020)

Key Responsibilities

  • Responsible for preprocessing brain MRI data to improve the performance of a brain tumor detection model (U-Net).

Project

1. MRI Brain Tumor Segmentation

Project Details
  • Key Achievements:
    • Applied and comparatively analyzed six preprocessing techniques, including Z-score normalization and White Stripe, on brain MRI data.
  • Insight:
    • Despite various preprocessing methods, performance improvement was minimal. A comparative analysis with the original data was conducted.
    • The analysis revealed that the delivered data was not the original and had already undergone some preprocessing.
    • This experience highlighted the importance of data originality and the preprocessing pipeline, contributing to the improvement of the company's data handling processes.

About

No description, website, or topics provided.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published