Think Autonomous

199€ or 2 x 99€

The LiDAR Object Detection DLC

Push your 3D Deep Learning skills to the limit, and master powerful 3D Object Detection architectures.

* Course restricted to Deep Point Clouds Engineers, enroll in the Deep Point Clouds course first

Enroll via Deep Point Clouds →Buy Now →
The LiDAR Object Detection DLC

Congratulations, you're at the Terminus!

This is the final step in your LiDAR Journey. The moment where you will push your skills to the limits, and learn to build ultra advanced 3D networks.

One thing has been missing all along...

In the Deep Point Clouds course, we're building hardcore fundamentals on 3D Deep Learning. We're learning how to use PointNet, and we're learning how to use Voxelization & 3D CNNs.

But all of it really is the "backbone" to more practical algorithms.

When you look around, you'll notice that engineers don't stick to using backbones and classifiers, they also learn to apply 3D Deep Learning in every day projects, like 3D Object Detection.

In LiDAR, 3D Object Detection is the most common and important application to master; and it was a missing element from my Deep Point Clouds course.

We have learned about classification, and it's now time to learn about object detection.

Just like with Computer Vision, most applications are based on object detection. Whether it's augmented reality, object recognition, drone tracking, autonomous driving, or any other; we must detect objects to understand our surroundings.

Self-Driving Car Companies heavily use LiDAR Object Detection

When you look at it this way; most companies rely on LiDAR. And most of these companies rely on object detection.

In fact, even non self-driving car companies who sell LiDARs and software use Self-Driving Car 3D Object Detection as example.

An example with Innoviz's Perception system:

Innoviz's Perception system

This is what this course is all about!

And I see this skill as the ultimate one to complete your journey in 3D Deep Learning.

These skills are valuable, valued, and hard to master. Going there alone would be pure madness, especially when a DLC exists and has everything already pathed for you.

This course is the only course you'll find on LiDAR Based Object Detection that teaches the blocks at this level. You may find videos explaining a few algorithms, or repositories showing how to run black boxes, but nothing will really help you "get it".

Until now...

LAUNCHING...

THE LIDAR OBJECT DETECTION DLC

Push your 3D Deep Learning skills to the limit, and learn to build advanced 3D Object Detection Architectures.

The LiDAR Object Detection DLC

Part I

Introduction to 3D Object Detection

We'll begin by understanding object detection, and directly zooming in 2 key object detectors, a point based and a voxel based approach. Zooming in early will help us understand some blocks, and make connections later in the course.

Intro to 3D Object Detection

Intro to 3D Object Detection

The 11 Most Common Blocks used in 3D Object Detection, and a brief overview of the most popular LiDAR Object Detection Algorithms

3D Bounding Boxes

3D Bounding Boxes

The difference between 2D and 3D Bounding Boxes, and what to consider when detecting objects in 3D (some objects may fly, be oriented differently, we'll see these cases too)

Point-RCNN Deep Dive

Point-RCNN Deep Dive

An in-depth exploration of the Point-RCNN algorithm (this algorithm is one of the top algorithms used in LiDAR Object Detection)

Apple's VoxelNet algorithm

Apple's VoxelNet algorithm

An exclusive article on VoxelNet, Apple's architecture to convert a point cloud into a voxel, run 3D CNNs, and generate bounding boxes.

Part II

3D Object Detection Deep Dive

We'll then dive in the advanced and state of the art architectures to generate bounding boxes.

PV-RCNN Thorough Dive

PV-RCNN Thorough Dive

A 6-Step Deep Dive into a State-Of-The-Art Architecture in 3D Object Detection: Point Voxels-RCNN

Keypoint Sampling

Keypoint Sampling

Learned and Traditional Approaches to Sample Point Clouds, and the ONE algorithm most state of the art algorithms use

Point-Voxel Fusion

Point-Voxel Fusion

A hardcore technical understanding of the fusion process between points and voxels, and how to "PointNet-ise" any voxel.

Keypoint Weighting

Keypoint Weighting

How to use Keypoint weighting to tell which points are foreground and background, and how this knowledge can be used to refine bounding boxes

Part III

Workshops & Implementations

In the final part, we'll see how to generate 3D bounding boxes. You'll have 2 workshops, on anchor box generation, and a complete architecture assembly.

3D Anchor Box Generation Workshop

3D Anchor Box Generation Workshop

Implement your own Anchor Box Generation algorithm, and test it to detect and track Flying Aircrafts.

Region Proposal Networks

Region Proposal Networks

The secret behind Region Proposal Networks and how to use Anchor Boxes in Neural Networks to generate Bounding Box Coordinates in 3D

Advanced Network Training

Advanced Network Training

3 Advanced Techniques to implement when training heavy modules that make up several modules (and how to avoid gradient issues)

Object Detection On Device

Object Detection On Device

How to run ANY 3D Object Detector algorithm in minutes, in a Jupyter environment, with active 3D Visualization

And many more, including:

And many more, including:

  • An advanced Research Paper Breakdown of 21+ Object Detection models (some of them include LiDAR-Camera "Deep Visual Fusion", which is an extra element I kept in the course)
  • How ROI Pooling works, and 2 techniques to implement it
  • How to build One Stage Object Detectors in 3D
  • How Gradient Scaling works, and why it's particularly important with Point Clouds
  • Where to find the best pretrained object detection weights
  • The secrets of CUDA programming, and how to use it with Point Clouds
  • and many, many more...

And obviously, we'll have a major workshop where you'll assemble your own 3D Object Detection Architecture

Here is a picture of what you'll do:

3D Object Detection workshop

FAQs

How long is the DLC and what do I get after it?

In this DLC, we will leverage what you learned in the Deep Point Clouds course. Therefore, we won't reexplain some fundamentals which are there.

Then, expect the content to take you 3-6 hours to understand.

After completing this course, you'll be a pioneer in the object detection world, and know more about it than 90% of the LiDAR Engineers.

Can I take this course if I already have 3D Deep Learning skills?

Currently, the only way to take this course is to be a client of the Deep Point Clouds course.

This is because these 2 courses work together, and following the DLC alone wouldn't really make sense.

Do I also have lifetime access?

Yes, just like with all my courses, this one comes with a lifetime access, and frequent updates.

☄️ How is this course unique?

"The only course on 3D Object Detection with Deep Learning"

This course is unique because in essence, it's the first and only course teaching object detection on LiDARs using Deep Learning.

Online, you can find a few tutorials or blog posts, you can find traditional approaches, but nothing about Deep Learning is really packed in a course that takes you from A to Z.

Having a course is important, because it helps connect the different bricks together, and gives you practical skills and experience rather than knowledge.

In fact, if I had to qualify this course, and the engineers who go through it using one adjective, it'd be... BOLD!

How bold should you be to take the course?

Let's see 3 levels of boldness:

  • Roller-Coaster: It's frightening but a well-defined and tested path.
  • Rock climbing: It's scary, requires more boldness, but still controlled.
  • Sky-Diving alone: Extreme level of boldness, you're on your own.

Because there aren't hundreds of courses on the topic, we'll be somewhere towards Level II and III.

If you're the kind who likes to explore uncharted waters, then this course will be for you.

3 levels of boldness: Roller-Coaster, Rock climbing, Sky-Diving
Mindmap of 3D object detection

Won't I be alone then?

No. And this is the point: it's a course! Its goal is to take you from A to Z. Which is why it will contain lots of fundamental elements, explanations, paper breakdown, tutorials, and even a mindmap explaining the entire field of 3D object detection.

At the end of this exploration, you'll have built practical skills in 3D Deep Learning, you'll know how to integrate 3d object detectors on almost any 3D scenario, and have a solid understanding of the research going on in this field.

DEEP POINT CLOUDS DLC

The LiDAR Object Detection DLC

The LiDAR Object Detection DLC

Push your 3D Deep Learning skills to the limit, and master powerful 3D Object Detection architectures.

199€

or 2 x 99€

* Course restricted to Deep Point Clouds Engineers, enroll in the Deep Point Clouds course first

Enroll via Deep Point Clouds Buy Now