Computer Vision: Instance Segmentation Project Demo

Hi there! My name is Andrei. I am a Junior Data Scientist/Machine Learning Engineer. In this video I am going to demonstrate a real-world Computer Vision project that I did in Practicum by Yandex, Data Science Bootcamp.
I found Instance Segmentation problem really interesting and challenging.
Check out my Medium article on the Instance Segmentation: https://medium.com/@andreibatomunkuev/instance-segmentation-most-challenging-problem-in-computer-vision-c8d2a37bce06

Follow me in:
- LinkedIn: https://www.linkedin.com/in/andrei-batomunkuev/
- GitHub: https://github.com/abatomunkuev
- Medium: https://medium.com/@andreibatomunkuev

Check out my portfolio for more information about me:
https://visual-portfolio-batomunkuev.vercel.app/




▬ Contents of this video ▬▬▬▬▬▬▬▬▬▬
00:00 - Introduction
00:32 - What is Instance Segmentation and why it is the most difficult task?
00:55 - Company that provided the task and data
01:13 - Project setup
02:06 - Data Overview
02:32 - Data Preprocessing for EDA
03:02 - Exploratory Data Analysis (EDA)
03:58 - Data Preprocessing for Deep Learning models
04:10 - Baseline model Mask R-CNN
04:25 - Creating a custom Dataset for Mask R-CNN
05:54 - Building a baseline model and training
06:38 - Baseline model evaluation and visualization
07:40 - Deep Learning Model Research: Detectron2
07:57 - Running pre-trained Detectron2 model
08:17 - Creating a custom Dataset for Detectron2
10:19 - Dataset visualization
10:30 - Building a Detectron2 model and training
10:49 - Detectron2 visualization
11:19 - Conclusion