Advanced in Computer Vision

Dr. Vu Anh Tuan

Week 1- Advanced Image Classification

Course introduction.

Recap of Convolutional Neural Network (CNN).

Introduction to advanced models such as VGGNet, ResNet, DenseNet, EfficientNet.

Lab: Setup TensorFlow / Colab.

Exercise (5%): Build VGG from scratch for a subset of images from ImageNet.

Assignment (20%): Build CNN models on a Vietnamese medical imaging data for eye diseases (courtesy of Cao Thang Hospital).

Week 2 - Object Detection

Introduction to object detection and advanced models such as YOLO and Faster RCNN.

Overview of recent advanced models such as DETR, FPN, NAS-FPN.

Lab: Collaborative annotation tool - MIT LabelMe / VGG Annotator

Exercise (10%): Train your own object detector with Faster RCNN on Vietnamese traffic

Week 3 - Image Segmentation

Introduction to image segmentation and advanced models such as Mask RCNN and U-Net.

Lab: Train your own segmentation model with mini FCN.

Exercise (10%) Davis’s challenge : single-frame object segmentation

Week 4 - Transfer Learning & Data Augmentation

Introduction to transfer learning.

Introduction to data augmentation: Elastic Transform & Augmentor

Data Augmentation tools: imaug, Albumentations, RandAugment.

Week 5 : Guest Lecture

Week 6 - Object Tracking

Lecture about object tracking : optical flow, sequence model, etc.

Introducing the final project on Video Object Segmentation (Davis Challenge @ CVPR 2020)

Exercise (10%): Build a baseline model 

Week 7 - Final Project Baseline Report

Project consultation with advisors.

Exercise (10%): write a baseline report (Introduction, Method, and Results) for your final project using Latex / Overleaf.

Week 8 - Guest Lecture

Project consultation with advisors.

Exercise (5%): update the report with new features and results.

Week 9 - Final Project Report and Presentation

Final Project (30%): project report submission and presentation.