Follow and star
never get lost
Institute of Computer Vision
Public ID｜Computer Vision Research Institute
Learning group｜Scan the QR code to get the joining method on the homepage
Computer Vision Research Institute column
Column of Computer Vision Institute
Since Transformer and BERT came out, they have begun to dominate the NLP field. Recently, Transformer-based multimodal large models and AIGC generation directions have also become research hotspots in the industry and academia. At the same time, large models are also widely used in data labeling and model distillation in autonomous driving.
Since Transformer and BETR came out, they have begun to dominate the NLP field. With the development of the past few years, Vision Transformer has gradually replaced the previous CNN structure on various visual benchmark datasets, and the overall architecture is simpler. Recently, Transformer-based multimodal large models and AIGC generation directions have also become research hotspots in the industry and academia. At the same time, large models are also widely used in data labeling and model distillation in autonomous driving.
How to get started?
The design idea of Visual Transformer is very different from the previous manual design of CNN. It is not very intuitive to understand, and it is not easy to learn by yourself. Many students often don't know how to start when they first learn. Most people don't know how to apply Vision Transformer to specific tasks, for example, how to design a specific model structure for perception tasks, and how to choose a more suitable Transformer according to their business needs. Models, how to apply large visual models to your own business or research tasks, etc.
Scan the code to learn the course
Closely combined with actual combat, it helps everyone better understand the implementation details of the algorithm, and systematically and comprehensively introduces the segmentation and detection model based on Transformer. This is the first complete visual Transformer theory and practical tutorial in China. It is especially suitable for beginners who are just getting started and students who need to engage in autonomous driving perception or image algorithms in business, and it is also very suitable for practitioners in other directions who want to systematically understand new directions.
The course outline is as follows:
main lecturer
Mr. Tiger, Doctor of Top2, currently a researcher in the industry. Current main research directions: image and video detection and segmentation, multimodal scene understanding, object tracking, and multimodal large models. He has published nearly 30 related papers in top computer vision conferences (CVPR, ECCV, ICCV, NeurIPS, ICLR, etc.)/top journals (T-PAMI, IJCV, TIP), among which 15 papers were published as the first author. Familiar with the design and implementation of commonly used segmentation and detection algorithms. Guided more than 6 junior doctoral and master students to publish top conference papers.
This course is suitable for everyone
Undergraduate/Master/PhD in the research direction of computer vision and autonomous driving perception;
2D/3D perception related algorithm engineers for autonomous driving;
Those who want to transfer to the Transformer-based perception algorithm;
Algorithm engineers and enterprise technical managers who need to be improved at work;
The basics required for this course
Have a certain foundation of python and pyTorch, and be familiar with some basic algorithms commonly used in deep learning;
Have a certain understanding of 2D perception, including basic solutions such as detection and segmentation;
A certain foundation of linear algebra and matrix theory;
The computer needs to have its own GPU, which can be accelerated by CUDA (at least 12GB of video memory);
After school harvest:
Able to have a systematic and in-depth understanding of recent Transformer-based segmentation and detection models.
Learn how to build your own Transformer perception system to solve some multimodal tasks.
Be able to proficiently implement some Transformer segmentation and detection algorithms, and master its algorithm improvement and application in the automatic driving system, to achieve code-level understanding.
After completing this course, you can carry out your own research work on Transformer segmentation and detection or design new methods in algorithm engineering.
Meet multi-industry practitioners and learning partners, and achieve a deeper understanding in the exchange.
start time
The class will officially start on September 8, 2023, join us to learn the basics, the class will end 2 months after the start of the class, offline teaching, Q&A in the WeChat group (the communication environment is very good, a very important part)!
course consultation
Scan the code to join learning!
Add a small assistant to consult and receive a course gift package!
(WeChat: AIDriver004)