Labeled point cloud files with 2 echos in .bag and .npy format.
25 bags, first 19 fully labeled and another 6 (numbers 20-25) without cutting out human. 

Folders:
/1_train_val_set:                Fully labeled, using accumulated start frames, also cut out people
/2_test_set:                     Labeled, using accumulated start frames, people still also labeled as smoke

Subfolders:
/numpy_files                     Contains .npy files
/rosbags                         Contains .bag files

Naming: 
<number>-dust_labeled_spaces     Dust bag, fully labeled, also cut out human using space information
<number>-smoke_labeled_spaces    Smoke bag, fully labeled, also cut out human using space information
<number>-smoke_labeled           Smoke bag, labeled, people still also labeled as smoke

.bag files:
Topics: /velodyne_points_dual and /velodyne_points
written at 10 Hz
Both topics are PointCloud2 messages
channels per topic: [x, y, z, intensity, ring, l_non, l_dust, l_smoke]

.npy files: 
shape: [number_frames, points_per_frame, 13]
13 entries: [x_echo_1, y_echo1, z_echo_1, intensity_echo_1, ring, x_echo_2, y_echo_2, z_echo_2, intensity_echo_2, ring, label_non, label_dust, label_smoke]
echo_1: corresponds to rostopic /velodyne_points_dual, echo_2: corresponds to rostopic /velodyne_points