What is the hottest direction of computer vision?
Of course it's target detection!
Object detection is one of the core issues in the field of computer vision. In the past two years, there have been hundreds of related papers in the world's top conferences, attracting more and more attention. Whether it is face recognition, automatic driving, text detection, human-computer interaction, it is inseparable from the cornerstone of target detection.
With the improvement of accuracy year by year, the landing of target detection technology has become more and more mature, and it has been widely used in various fields such as industry, transportation, remote sensing, medical treatment, entertainment and so on!
As the "China AI Head Goose", Baidu has accumulated rich experience in target detection algorithm research, industrial practice, and industrial implementation!
In July 2020, the Baidu Paddle Detection R&D team explored and open-sourced the industrial-grade target detection model PP-YOLO. The accuracy and prediction speed are better than YOLOv4, and it is in the leading position in the industry!
In August 2020, the Baidu Vision team performed well in ECCV 2020, the top international competition in the visual field, and won the "Double Champion" of Tiny Object Detection and COCO in the field of object detection!
In order to help more developers deeply understand the field of target detection, the Baidu Vision team and the Paddle Detection R&D team have joined forces, combined with years of practical experience, taught in person, and led everyone to specialize in target detection!
Download and install command ## CPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /cpu paddlepaddle ## GPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /gpu paddlepaddle-gpu
7 days to conquer target detection
- Top famous teachers teach
Global ECCV target detection double champion team teaching and research, Paddle Detection senior research and development teaching, based on years of top conference scientific research and 100+ AI industry landing experience, the whole process of live teaching.
- Deeply dig through the algorithm principle
Detailed dismantling of RCNN, YOLO, and AnchorFree series algorithms, as well as the industrial-grade detection model PP-YOLO with SOTA performance, will take you through the fog step by step!
- Real industrial project combat
In-depth analysis of 4 real industrial cases featuring Baidu's characteristics: road sign inspection, industrial quality inspection, power inspection, and intensive commodity inspection in commercial supermarkets.
- Hand in hand to play the latest CVPR2020 competition
The 2019 CVPR target detection world champion, the first time to publicly share the champion cheats, with the latest competition in 2020, dense object detection based on the COCO data set!
special gospel
-
Excellent students have the opportunity to join Baidu
-
Certificate of Completion by mail
-
In-depth communication with Baidu senior R & D throughout the process
-
Outstanding open source ecological contributors will be prioritized to become PPDE ( Paddle Paddle Technology Expert)
-
The whole course is free, including live broadcast, practice, homework correction, Q&A, etc., and free GPU resources are provided
Course Outline
01A review of target detection tasks
1. Development history of target detection, important basic concepts (bbox, nms, IOU, mAP)
2. Dataset format introduction
3. Overall introduction and use of Paddle Detection
【Combat】Road Sign Detection
02 Detailed explanation of RCNN series target detection algorithm
1. Introduction of RCNN series models
2. Detailed explanation of FasterRCNN module
3. Paddle Detection training RCNN model
03 RCNN series algorithm optimization strategies and cases
1. Introduction to FPN
2. Introduction to CasacdeRCNN
3. Two-stage model optimization for speed and accuracy
4. Industrial quality inspection cases
[Actual combat] Defect detection of industrial quality inspection based on RCNN
04 Detailed explanation of YOLO series target detection algorithm
1. Introduction of YOLO series
2. Detailed explanation of YOLOv3 module
3. Train the YOLO model
05 PP-YOLO optimization strategy and case sharing
1. Introduction of PP-YOLO
2. Optimization of YOLO model speed and accuracy
3. Electric power inspection case
[Actual combat] PP-YOLO industrial quality inspection defect detection
06 Introduction and practice of AnchorFree series algorithms
1. AnchorFree series introduction
2. Introduction to FCOS
3. Introduction to CornetNet/TTF-Net
【Combat】AnchorFree Industrial Quality Inspection Defect Detection
07The world champion takes the top of the game
1. Announce the latest 2020 CVPR competition questions
2. The world champion shares the secrets of the 2019 competition
【Actual Combat】Commercial Supermarket Intensive Commodity Detection
(Dense Object Detection)
Instructors
Teacher Qingqing: Baidu senior R&D engineer
Responsible for the visual direction of the flying paddle , engaged in the research and development of the flying paddle framework and visual algorithms for many years, participated in the research and development of the flying paddle system and visual model library from 0 to 1, and was responsible for the research and development of the flying paddle classification, segmentation, target detection, generation and other models, and the model He has rich experience in system optimization and application of vision algorithms in the research and development of compression tools.
Mr. Pengpeng: Baidu senior R&D engineer
The main author of the server-side PP-YOLO model, the core R&D backbone of the Paddle Detection target detection library, has an in-depth understanding and research on the principles and optimization of the YOLO series of models, and has developed and released a variety of server-side and mobile-side YOLOv3 models. The speed has reached the forefront of the industry.
Mr. Guan Guan: Senior R&D Engineer at Baidu
The core R&D backbone of Paddle Detection, a flying paddle target detection suite , has built an end-to-end development kit for flying paddle target detection from 0 to 1. The developed detection model has reached the forefront of the industry, with years of algorithm accumulation and practical experience in target detection. In-depth understanding of the detection field.
learning motivation
Worried about not being able to keep learning? Don't worry, you need encouragement along the way.
We have set up a wonderful competition PK link, and we have prepared 100 exquisite gifts waiting for you to come and get them!
Baidu Flying Paddle Official Certification Completion Certificate
Scan the code to pay attention to the [ Flying Paddle ] public account
Re: Championship Team
Get a dedicated registration channel
Download and install command ## CPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /cpu paddlepaddle ## GPU version installation command pip install -f https://paddlepaddle.org.cn/pip / oschina /gpu paddlepaddle-gpu