vadbeg_projects.txt

1. News Scrapper
  GitHub: https://github.com/Vadbeg/NewsScrapper
  Description:
    Project for reading news from rss files. Includes command line utility and web application. 
    The main thing about this project is fasts, flexible and easy acces to news in format you want.
    You can export every article into fb2, pdf and html.
  Languages:
    1. Python
  Technologies:
    1. Django
    2. SQL
    
2. Shoe size identifier using ML
  Description:
    Using photo/video from mobile phone camera you can identify your shoe size. On first stage
    was written algorithm wihch used basic Computer Vision tools and clusterization. 
    This algorithm was used for data markupping. Next, marked-up data was used for
    training UNet. UNet works faster and better than clustering in this task.
  Languages:
    1. Python
  Technologies:
    1. PyTorch
    2. OpenCV
    
3. Estimating sells of retail goods (Kaggle)
  Description:
    Using data about sales, sell prices, calendar data and sells for every day of 2 year-period
    you need to identify cells for next 28 days. My algorithm uses seq2seq Neural Network based on
    LSTM blocks.
  Languages:
    1. Python
  Technologies:
    1. PyTorch
    
4. Handwritten grapheme classification (Kaggle)
  Descrition:
    On basics of photoes of graphemes I need to identify which graphemes are presented on
    image. Any grapheme can consist of 3 parts (root, vowel diacritics, consonant diacritics).
    I used seresnext50 model for this task. Mine accuracy on this task: 0.9186.
  Languages:
    1. Python
  Technologies:
    1. PyTorch
    2. torchvision
    3. OpenCV