Thats another great video!Thanks One thing I am struggling with is whenever I increase the octree level in connected component algo/tool like if I go beyond 9 it is crashing. is it system limitation or version specific issue?
Hey there! I'm thrilled you enjoyed the video. That's fantastic. Increasing the octree level beyond 9 can lead to crashes. This is likely due to memory limitations. It's not usually version-specific. Think of it like this: each level drastically increases the number of cells in your octree. This explodes memory requirements very quickly. Your system might be running out of RAM. Processing larger point clouds or increasing the octree depth requires substantial memory resources. What are your system specs? Knowing your RAM, CPU, and GPU can help diagnose this. It’s like trying to fit a huge puzzle into a small box. If the puzzle's too big (your octree), it won't fit in the box (your RAM).
Hi Florent, thank you for the great video. I am dealing with underwater vegetation lidar data, which is messy indeed. I'd like to differentiate between marine algal canopies and the 'true' bottom of the sea, even though they are classified together within a bottom layer class. Do you have some point cloud labelling examples online that may be more suitable for such a situation? For example, detecting planes with the ransac shape detection was quite difficult to interpret for such an endeavour. Thank you again, and any input will be helpful!
Hey there! Thanks for reaching out. I saw your struggles with underwater vegetation LiDAR data. It's definitely tricky stuff. 🌊🌿 I understand you want to separate algal canopies from the actual seabed, even when classified together. This is a common challenge. Plane detection with RANSAC can be hard to interpret in such environments, I agree. Let's think outside the box. Density-based clustering could be a game-changer. Imagine grouping points based on how close they are. This will help you differentiate the denser canopy from the sparser seabed. Another idea? Explore point features like intensity and return number. Canopy and seabed often reflect laser differently. These features can be your secret weapon. For labeling, consider creating custom labels specific to "canopy" and "seabed". This allows for more accurate training of machine learning models down the line. My 3D Segmentor Course Line dives into point cloud processing workflows that can be very useful + advanced stuff like segmentation, recognition, classification, and 3D modeling. I hope this helps!
@@FlorentPoux Thank you for the robust response! I will speak with my employer, I hope to take your course ASAP :) I know how helpful it would be! Thanks as always Florent
Another great video, Florent! Thank you!
Thanks again!
Your content you are teaching us is not good. Excellent man !!!
Thanks!
Thank you very much! We are waiting for the next step (training an algorithm). 🎉
on the todo :)
Great video!! Looking forward to the tutorial on fully automated workflow using Python.😊
Coming soon!
Just what I was looking for, thanks Florent!
Haha beautiful! Happy to deliver!
@@FlorentPoux Me too :)
Whats the next step, im trying to build a autonomous drone for my school project so this is really helpful🎉.
Beautiful! Super happy this helps you! The next step is to train a supervised model with the dataset for your drone recognition capabilities
Thats another great video!Thanks
One thing I am struggling with is whenever I increase the octree level in connected component algo/tool like if I go beyond 9 it is crashing. is it system limitation or version specific issue?
Hey there! I'm thrilled you enjoyed the video. That's fantastic.
Increasing the octree level beyond 9 can lead to crashes. This is likely due to memory limitations. It's not usually version-specific. Think of it like this: each level drastically increases the number of cells in your octree. This explodes memory requirements very quickly. Your system might be running out of RAM. Processing larger point clouds or increasing the octree depth requires substantial memory resources.
What are your system specs? Knowing your RAM, CPU, and GPU can help diagnose this. It’s like trying to fit a huge puzzle into a small box. If the puzzle's too big (your octree), it won't fit in the box (your RAM).
Hi Florent, thank you for the great video. I am dealing with underwater vegetation lidar data, which is messy indeed. I'd like to differentiate between marine algal canopies and the 'true' bottom of the sea, even though they are classified together within a bottom layer class. Do you have some point cloud labelling examples online that may be more suitable for such a situation? For example, detecting planes with the ransac shape detection was quite difficult to interpret for such an endeavour. Thank you again, and any input will be helpful!
Hey there! Thanks for reaching out. I saw your struggles with underwater vegetation LiDAR data. It's definitely tricky stuff. 🌊🌿
I understand you want to separate algal canopies from the actual seabed, even when classified together. This is a common challenge. Plane detection with RANSAC can be hard to interpret in such environments, I agree.
Let's think outside the box. Density-based clustering could be a game-changer. Imagine grouping points based on how close they are. This will help you differentiate the denser canopy from the sparser seabed.
Another idea? Explore point features like intensity and return number. Canopy and seabed often reflect laser differently. These features can be your secret weapon.
For labeling, consider creating custom labels specific to "canopy" and "seabed". This allows for more accurate training of machine learning models down the line.
My 3D Segmentor Course Line dives into point cloud processing workflows that can be very useful + advanced stuff like segmentation, recognition, classification, and 3D modeling.
I hope this helps!
@@FlorentPoux Thank you for the robust response! I will speak with my employer, I hope to take your course ASAP :) I know how helpful it would be! Thanks as always Florent