李滢滢

> 您可以按照Stanford3dDataset_v1.2的样式进行数据学习。如作者在Readme.txt中所述，然后利用collect_indoor3d_data.py进行数据重组，并使用gen_indoor3d_h5.py生成HDF5文件。 小仙女还在吗，我搞不出来正确的h5文件，想哭

> 1. Set FLAGS.normal = True in train files. > 2. Replace > l0_xyz = point_cloud > l0_points = None > > by > l0_xyz = tf.slice(point_cloud, [0,0,0], [-1,-1,3]) >...

I have encountered that, you open the terminal and enter nvidia-smi to see if the graphics card driver is still there.Maybe you will find that graphics card driver is lost.

Yeah,I have solved it. But after such a long time,I don't remember the reason,sorry. 发自我的iPhone ------------------ Original ------------------ From: Qimin Chen ***@***.***> Date: Tue,May 4,2021 11:05 AM To: charlesq34/pointnet2 ***@***.***>...

> 1. Set FLAGS.normal = True in train files. > 2. Replace > l0_xyz = point_cloud > l0_points = None > > by > l0_xyz = tf.slice(point_cloud, [0,0,0], [-1,-1,3]) >...

解决了 发自我的iPhone ------------------ Original ------------------ From: TongLiu

> 1. 在训练文件中设置FLAGS.normal = True。 > 2. 替换 > l0_xyz = point_cloud > l0_points =无 > > 通过 > l0_xyz = tf.slice（point_cloud，[0,0,0]，[-1，-1,3]） > l0_points = tf.slice（point_cloud，[0,0,3]，[-1，-1， 3]） > > 在模型文件中。...

> #41 I already understand about how the code works now.Thanks again!

> 我的点云数据只有x，y，Z，强度值这四个维度。如何训练？您能帮我吗？战车 hi,did you work it out ?

> 你好！我得到了一个具有pointnet_cls_ssg.py的模型，当我用256个输入点进行评估时，我得到了58％的准确度，这与85％以上的论文不一样？ why don't use 1024?