A really good explanation only focusing on the important stuff. You got my thumbs up for this one bro.
@Fallen___Ash
2 жыл бұрын
Thank you for this informative video. You explained all the important parts of the paper in simple terms using examples and I really appreciate it
@chorhanchean1175
5 ай бұрын
Straight to the point, thank you!
@keremkurban8156
Жыл бұрын
I wonder if they just tried 1024 feature space at first for segmentation and when the performance was not very good, if they said 'well lets just add the features from before' :D I would like to try this myself but this also raises the question, if the problem was more complex, would we need to make the network deeper and concatanate 3,4,5 feature tensors than 2 shown here?
@hussainmujtaba638
Жыл бұрын
Subscribed, amazing explanation. Hope to get more content from you ahead.
@pierre-adriencarnelutti6099
Жыл бұрын
great explanation and going to the point, Thank you very big!
@mrdigs7919
Жыл бұрын
just wow, amazing content
@chidambarhunakunti6590
3 жыл бұрын
Thank you soo much
@amitgupta-zu9cs
2 жыл бұрын
Thanks for great explanation. You made life simpler.
@sharatgujamagadi6809
5 ай бұрын
good one :)
@kulkarniajinkya
Жыл бұрын
Great video, thanks for sharing. Are you from Germany by any chance?
@TheBhizmi
Жыл бұрын
Perfect. Thank you!!!
@chidambarhunakunti6590
3 жыл бұрын
Can you please do such a video on pointNet++ as well?
@MinimalDebug
3 жыл бұрын
It's in the workings..
@felixwhise4165
2 жыл бұрын
@@MinimalDebug how is the process on that? :D
@adityawagh
Жыл бұрын
@@felixwhise4165 Check the channel, there's a video on PointNet++ and VoxelNet
@alejandrogm7617
14 күн бұрын
@@adityawagh where did the PointNet++ video go?
@zhurui7845
2 жыл бұрын
Good explanation, it helps me, thanks!
@j________k
Жыл бұрын
Great video! One question with the combined global and local features do the segmentation categories exist globally that is to say.... Would global class chair have segment 0 leg, 1 arm, 2seat and global class plane have segment 0 wing, 1 engine etc... Or do the segment labels have to be universally unique. Otherwise it would see that the number of segments would have to be massive compared number of global classes
@rishavkumarparamhans1595
2 жыл бұрын
Thanx for this! it was really helpful
@naziahossain3950
2 жыл бұрын
thanks a lot for the explanation
@mingjuhe773
2 жыл бұрын
Thanks, you are great.
@angokenseifan
2 жыл бұрын
good explanation
@randywelt8210
2 жыл бұрын
12:50 I think ordered pointcloud means it contains always the same amount of points filled up with Nan values.
@randywelt8210
2 жыл бұрын
OK know I read it. Ordered in ur case means what u explained it different from what it means for PCD files.
@jasorsighosh7838
2 жыл бұрын
ordered means there is a deterministic bounded way to figure out the neighbour for any random point.
@andrzejreinke
2 жыл бұрын
this is good !
@m1a2tank
2 жыл бұрын
do you think point works well with 2d vector image? i think it can be alternative for the cnn which get attack by rasterization error.
Пікірлер: 29