vLAR-group
train my datasets
Hello, thank you for your work. I would like to ask why there are large black areas in the visualization results when I use my own dataset (ancient building data) to construct initial superpoints. I think it may be unrecognized areas. So, for the hyperparameters in VCCS and region growing algorithms, how should I adjust them to achieve the expected results? Looking forward to your prompt reply!
hi @lrnb12345 ,
Could you provide some visualization examples for your personal data? For the region growing+vccs, you may try to set a larger threshold for normal distance on region growing if there are too many invalid areas. The smooth and curvature threshold in https://github.com/vLAR-group/GrowSP/blob/b7eaece0847189c7b94854a03d7a3816bf478236/data_prepare/initialSP_prepare_S3DIS.py#L58
Thank you for your reply. After debugging and modifying the parameters overnight before your reply, I did solve the problem and displayed the appropriate superpoint area on my dataset. However, I found that you did not provide the visualization code for the results. Could you please analyze the visualization code for me? I would greatly appreciate it！
------------------ 原始邮件 ------------------ 发件人: "vLAR-group/GrowSP" @.>; 发送时间: 2024年3月27日(星期三) 晚上9:57 @.>; @.@.>; 主题: Re: [vLAR-group/GrowSP] train my datasets (Issue #17)
hi @lrnb12345 ,
Could you provide some visualization examples for your personal data? For the region growing+vccs, you may try to set a larger threshold for normal distance on region growing if there are too many invalid areas. The smooth and curvature threshold in https://github.com/vLAR-group/GrowSP/blob/b7eaece0847189c7b94854a03d7a3816bf478236/data_prepare/initialSP_prepare_S3DIS.py#L58
— Reply to this email directly, view it on GitHub, or unsubscribe. You are receiving this because you were mentioned.Message ID: @.***>