Language: en | ru

Portfolio

Computer vision / AI

1. Passport scan recognition using neural networks
2. AI system for tracking user emotions with web-camera
3. AI capable of learning to play Ping-Pong game

Passport scan recognition using neural networks

2022-12

• github.com/d-01/passport-ocr-test
• Recognition is done in three steps:
  1. Passport's page localized on an input image
  2. Textual fields detected in the target area
  3. Individual characters are recognized with three different methods (OCR, API, neural networks)
• No installation required, runs entirely in Google Colab (open).
• Tech-stack: python, opencv, pillow, OCR, tensorflow, keras, numpy, cnn

AI system for tracking user emotions with web-camera

2022-01

• github.com/d-01/webcam-emotion-recognition
• No installation required, runs entirely in Google Colab (open).
• Uses web-camera (browser access required), tracks face in the frame and recognizes emotion.
• Tech-stack: python, tensorflow, keras, opencv, cnn, mobilenet, fer, facial expression recognition, valence-arousal, face detection, dlib, google colab, javascript, html canvas

AI capable of learning to play Ping-Pong game

2021-09

• Open Jupyter Notebook
• Video
• Training neural network to play Atari Pong game using raw-pixel data as input.
• Tech-stack: RL, reinforcement learning, tensorflow, gym, Q-learning, CNN, DQN, replay buffer