Відео

very clear explained，thanks

Reinforcement learning lecture is coming?😊

These are life savers

nice!

Amazing work! Hats off to you good sir, concise explanation of a very complex topic

Hallo, are the slides for this lecture available online.

is Graph slam only full (offline) slam or can it be online too?

where can I find the lecture note?

How do I preprocess the NGSIM dataset and implement vehicle trajectory prediction in Python?

Thank you for sharing sir. I have a question: why is the complexity O(k^2.4 + n^2)? I understand that O(k^2.4) comes from the matrix inversion, but where does O(n^2) come from? Why isn't it something like O(n^2 * k + n * k^2) due to matrix multiplication?

As a prospective PhD this is a rare gem and the channel in its entirety !!!

What is the size of the model that performs this? How fast does it run?

The time when 3D data will be actively used in generative models seems to be approaching.

Excellent 👌

Too good - I wish I had studied your videos 10 years ago when I was starting out. Somehow, the books don't give an intuitive picture making this a much more difficult area to approach than it should be. Prof. Stachniss, you should write a book with some good pen and paper and programming exercises. Forstner is probably the best right now. (I work in self driving cars, on BEV modelling, and LOVE this subject).

Thank you, amazing explanation. 👍

An outstanding lecture on template matching both in theoretical and in pratical.

So is the initial input to both the KF and EKF the gaussian pdf functions or what else?

Very your explanation! Thank you! You did not mentioned accuracy in your video, could you explain how accuracy is calculated in this kind of model? The math behind isn’t clear for me.

Impressive work, could you please provide GitHub repository for reproducibility?

There are unknown way to visualize subspace, or vector spaces. You can stretching the width of the x axis, for example, in the right line of a 3d stereo image, and also get depth, as shown below. L R |____| |______| TIP: To get the 3d depth, close one eye and focus on either left or right line, and then open it. This because the z axis uses x to get depth. Which means that you can get double depth to the image.... 4d depth??? :O

I wish to apply this practically. I have only thought of the graph implementation.

Really nice ! What is the cheapest LiDaR sensor chip for 3D indoor Mapping ?

TNQ

on a slide 18 (computing alpha_t). isn't alpha_t = argmin(alpha) sum(pa(alpha)) ?

Great Work👍

So clearly explained! Thank you.

Thank you Professor.

Great work, Matteo! Cheers from Boulder

Hello, I enjoyed watching your video. And I have one question. In radar mapping, if you use distance thresholding like that, wouldn't newly observed objects not be mapped?

No

Amazing

I noticed LiDAR would most likely be something that is installed on top of a passenger car like a 'TAXI' sign, but that kind of appearance would not aesthetically compromise a TaxiBot design value. Couple of missing pieces in achieving autonomous level-5 practicallity would probably be : • multi weather urban road application under various kind of debris flying around the LiDAR 3D scope detection while in motion, and (but not limited to) • Black Box auto-pilot system capturing these events as & when the car is mobile on the road without time limiting factor on data memory captured. Do you reckon that a level-5 autonomous TaxiBot well equiped with Cameras + Radars + LiDARs + Black Box (w/ unlimited data time captured via cloud) ready to be rolled out on Chinese road (if citing for example Huawei level-5 software) ?

Thank you for this video.

I would be glad if s.o. has knowledge replies; should not we take the Jacobian of f w.r.t control signal u to get B_jacobian and take into account for adding the value B_Jacobian*u for calculating prediction of states ? At time 37:20 ? thanks.

my most productive 5 min today

This is an amazing lecture, thank you!

So is translation still linear, or also nonlinear like rotation

correction - Viola & Jones isn't from the nineties, the initial version is from 2001. anyways great lecture!

how to applay it in a case of mixed outliers/outliers-free measurements ?

This video is amazing. Thanks for clear and concise explanation. Watched with pleasure

May I know which sensors were used. Although the ARS-408 2D radar sensor was mentioned as an example there is nothing about the 3D radar sensor.

So much wisdom in these videos. Bedtime entertainment,

thank you , that's amizing , do you have dataset

nice!

big up cyrill

Also we use the nonlinear function equation for computing the prediction state values around the linearization point right? In every iteration we put the estimated values inside the nonlinear function as to compute the state to use them inside the prediction vector which is the first step of EKF right? Thank you

How did you fine tune params of Kiss-ICP on Mulran Kaist 03 ? i launched several times kiss-icp on this dataset and the odometry was awful. At high speed of the car the scan matching did not work at all.

😍