Tensorrt based in-depth learning model deployment practice tutorial!

Application background

As early as far away 1989 year , One is called ALVIVN For the first time, the company has used neural networks in automobiles , Carry out lane line detection and ground segmentation . today , Deep learning has been applied in many branches of auto drive system . The first is the field of perception , Commonly used sensors are cameras 、 Lidar and millimeter wave radar . Deep learning uses two-dimensional images or three-dimensional point clouds as input , Detect the obstacles 、 distinguish 、 Division 、 Tracking and ranging .

Mask RCNN

M3DSSD: Monocular 3D Single Stage Object Detector

PointPillars: Fast Encoders for Object Detection from Point Clouds

The second is the positioning field , Automatic driving usually requires centimeter level positioning accuracy , This makes the traditional high-precision map unreliable in many scenarios . In recent years, some methods use online map learning , Based on on-board sensor observation , Build high-definition maps dynamically , It is a more scalable way than traditional pre annotated HD maps , Provide semantic and geometric priors for autonomous vehicles .

HDMapNet: An Online HD Map Construction and Evaluation Framework

Thirdly, forecast planning , Using deep learning method can better predict the trajectory of obstacles , There are even ways to put perception - Combine forecasting to solve problems .

PnPNet: End-to-End Perception and Prediction with Tracking in the Loop

Because deep learning algorithm is widely used in automatic driving , This makes model deployment engineers hot , Most companies require Algorithm Engineers to design algorithms , Deploy migration again , Talents with both skills have always been the preferred candidates for automatic driving companies , The corresponding salary is also very considerable .

How to get started learning ？

Whether it's classification 、 Detect or split tasks , Most deep learning algorithms have corresponding open source implementations , Usually based on python, However, how to deploy the model to C/C++ In Engineering No systematic tutorials , As a result, many children's shoes stay on the optimization of the model , For this reason, we have launched the first industrial practical training course in China ：《 Deployment of deep learning model in autopilot 》, First line Algorithm Engineers lead the team ,10 Years of engineering experience , Help everyone make rapid progress , Little detours .
At present, most autopilot hardware platforms adopt NVIDIA's GPU programme , No matter on the industrial computer platform , Or embedded platform . Software aspect ,Nvidia Provides a wealth of SDK To help developers use GPU, For example, for model deployment TensorRT, Suitable for data processing NPP and NvMedia etc. , also cuDNN and cublas And so on , Even available cuda c More convenient to expand . This course mainly explains how to use TensorRT Model deployment , And the use of NPP and cuda Data preprocessing , among TensorRT and cuda Are the two key points of this course ,cuda Is the relative difficulty . The syllabus is as follows ：

About Instructor

The lecturer of the course is a master who graduated from Harbin Institute of technology , Yes 10 year C++ Programming experience ,2018 In, we began to contact the deployment of deep learning model ,2019 Joined a large automatic driving company in . Yes TensorRT and GPU Deep understanding of parallel programming . This course will focus on the deployment of in-depth learning model , Lead you to master TensorRT The use of and cuda Programming .

Highlights of the course

This course will adopt the idea of combining theory with practice , First of all, TensorRT Programming model and GPU/cuda To explain the relevant knowledge , Lead everyone to know why ; After that, the course will use classification 、 testing 、 Split three examples to show the detailed programming process , And give the relevant code , Achieve real industrial level sharing .

The harvest after learning

1. Be familiar with the preprocessing of model deployment 、 Post processing flow ;

2. Yes TensorRT Have a deep understanding of the overall structure of the frame , Be able to implement custom layer ;

3. Master classification 、 testing 、 The deployment process of the split model , Ability to deploy product models into hardware ;

4. Requirements for enterprise level post technology stack ;

Course requirements and object orientation

1. Computer vision related direction , Have a certain amount of Python perhaps C/C++ The basis of programming ;

2. It needs to be self provided GPU Or online rental server ;

3. Undergraduates who have a strong interest in the field of autonomous driving 、 Graduate student , Personnel engaged in model deployment .

Class time

2021 year 12 month 11 Classes officially open on , It took two months , The detailed schedule is subject to the announcement in the learning group .

Course consultation and purchase

▲ Code scanning learning course

Group number ：910070197
▲ Course consultation QQ Group , Learn more about

▲ Long press plus customer service wechat , Consult more