Custom Deep Learning.

[Felipe Paranhos]

  I am following the LIDAR360 version 8.1 manual to perform supervised custom classification using deep learning. 

First, I read through the manual, starting from page 303 (file:///C:/Program%20Files/GreenValley%20Suite/LiDAR360/8.1.1.0/doc/LiDAR360UserGuide-EN.pdf). 

My main issue is with the "Custom Deep Learning" process, specifically during the model training phase. 

The models I am training keep producing errors. 

I suspect that one contributing factor is that the images in the manual, particularly on page 305, do not match those in version 8.1.  

I found this video to help, but I have another question: [https://www.youtube.com/watch?v=pDtx0FnBNqw].

In the video, I couldn't fully understand this text: "The recommended ratio between training and validation data is 8:2 or 7:3, so you need to split the labeled samples into two parts beforehand. Theoretically, the more labeled data used for training, the better the quality of the trained model. It is also advisable for the sample data to cover all the scenes to be classified

[Solaire Guo]

Dear Sir/Madam,

The video you sent appears to be a tutorial for LiDAR360MLS. Could you please confirm whether you are using LiDAR360 or LiDAR360MLS? These are two different software programs.

For your reference, I have included a tutorial video for the Custom Deep Learning feature in LiDAR360. I hope this will be helpful to you.

LiDAR360 V8 - Classification | Point Cloud Custom Deep Learning LiDAR Software Tutorial

Best regards,