pointnet++的pytorch實現

代碼參考：https://blog.csdn.net/weixin_39373480/article/details/88934146

---

 

 

def recognize_all_data(test_area = 5):
    ALL_FILES = getDataFiles('/home/dell/qcc/pointnet/code/Pointnet_Pointnet2_pytorch-master/data_utils/data/indoor3d_sem_seg_hdf5_data/all_files.txt')
    room_filelist = [line.rstrip() for line in open('/home/dell/qcc/pointnet/code/Pointnet_Pointnet2_pytorch-master/data_utils/data/indoor3d_sem_seg_hdf5_data/room_filelist.txt')]
    data_batch_list = []
    label_batch_list = []
    for h5_filename in ALL_FILES:
        data_batch, label_batch = loadDataFile('/home/dell/qcc/pointnet/code/Pointnet_Pointnet2_pytorch-master/data_utils/data/' + h5_filename)
        data_batch_list.append(data_batch)
        label_batch_list.append(label_batch)
    data_batches = np.concatenate(data_batch_list, 0)
    label_batches = np.concatenate(label_batch_list, 0)

ALL_FILES：24個.h5格式的點雲文件,每個.h5文件包含1000個場景（最后一個.h5文件包含585個場景，共有場景23585個），每個場景有4096個點雲及其對應標簽。
其中18~29場景可視化如下圖（點雲數據本身包含有RGB顏色信息）：

第18個場景：

room_filelist： 室內場景共6個區域的23585個點雲文件。