csikasote/aims2020_visualrecognition

AIMS 2020, class on Visual Recognition

Visual Recognition from Images and Videos

This series of lectures will focus on visual recognition. In particular, we will dive into the state-of-the-art 2D object recognition models from single images, then study how to build deep learning models that recognize from sequence of frames instead of single images and finally we will talk about how we can extend these models for 3D object understanding.

Each lecture will consist of

• a video lecture: Normally, I would give the lecture live but due to the 9hr time difference with the west coast we will instead rely on Justin Johnson's video lectures for the EECS 498-007 / 598-005 Deep Learning for Computer Vision at University of Michigan.

• a Q&A: After you have watched the video lecture, I will hold a Q&A session where I will be answering questions regarding the material covered in the lecture. There you can ask me any questions or for clarifications.

• a test: After the Q&A session, I will give you a small test of multiple-choice questions regarding the material covered in the class. This test should take no more than 30min to answer.

• a lab: Finally, we will have a practical lab session where we will get hands on experience with train and using visual recognition models. After each lab session, I will ask you to write a short report answering a variety of questions. You will be graded based on the test and the lab report.

Lecture 1: 2D Object Recognition

In the first lecture, we will cover 2D object recognition from images. We will focus on the tasks of object detection, semantic segmentation and instance segmentation.

• LECTURE1.md: Reading material and the video lectures
• LAB1.md: Lab session and short report assignment (deadline: Sunday, May 3)
• TEST1.md: Test for lecture 1 (deadline: Tuesday, April 28)

Lecture 2: Video Recognition

In the second lecture, we will cover video understanding.

• LECTURE2.md: Reading material and the video lectures
• LAB2.md: Lab session and short report assignment (deadline: Sunday, May 3)
• TEST2.md: Test for lecture 2 (deadline: Thursday, April 30)

Lecture 3: 3D Object Understanding

In the third lecture, we will cover 3D object understanding.

• LECTURE3.md: Reading material and the video lectures
• LAB3.md: Lab session and short report assignment (deadline: Sunday, May 3)
• TEST3.md: Test for lecture 3 (deadline: Sunday, May 3)

Schedule

The schedule for the class can be found in SCHEDULE.md