def test_SemanticKITTISCN():
from xmuda.data.utils.visualize import draw_points_image_labels, draw_bird_eye_view
preprocess_dir = '/datasets_local/datasets_mjaritz/semantic_kitti_preprocess/preprocess'
semantic_kitti_dir = '/datasets_local/datasets_mjaritz/semantic_kitti_preprocess'
# pselab_paths = ("/home/docker_user/workspace/outputs/xmuda/a2d2_semantic_kitti/xmuda_crop_resize/pselab_data/train.npy",)
# split = ('train',)
split = ('val',)
dataset = SemanticKITTISCN(split=split,
preprocess_dir=preprocess_dir,
semantic_kitti_dir=semantic_kitti_dir,
# pselab_paths=pselab_paths,
merge_classes=True,
noisy_rot=0.1,
flip_y=0.5,
rot_z=2*np.pi,
transl=True,
bottom_crop=(480, 302),
fliplr=0.5,
color_jitter=(0.4, 0.4, 0.4)
)
for i in [10, 20, 30, 40, 50, 60]:
data = dataset[i]
coords = data['coords']
seg_label = data['seg_label']
img = np.moveaxis(data['img'], 0, 2)
img_indices = data['img_indices']
# pseudo_label_2d = data['pseudo_label_2d']
draw_points_image_labels(img, img_indices, seg_label, color_palette_type='SemanticKITTI', point_size=1)
# draw_points_image_labels(img, img_indices, pseudo_label_2d, color_palette_type='SemanticKITTI', point_size=1)
# assert len(pseudo_label_2d) == len(seg_label)
draw_bird_eye_view(coords)
def test_A2D2SCN():
from xmuda.data.utils.visualize import draw_points_image_labels, draw_bird_eye_view
preprocess_dir = '/datasets_local/datasets_mjaritz/a2d2_preprocess'
split = ('test',)
dataset = A2D2SCN(split=split,
preprocess_dir=preprocess_dir,
merge_classes=True,
use_image=True,
noisy_rot=0.1,
flip_y=0.5,
rot_z=2*np.pi,
transl=True,
fliplr=0.5,
color_jitter=(0.4, 0.4, 0.4)
)
for i in [10, 20, 30, 40, 50, 60]:
data = dataset[i]
coords = data['coords']
seg_label = data['seg_label']
img = np.moveaxis(data['img'], 0, 2)
img_indices = data['img_indices']
draw_points_image_labels(img, img_indices, seg_label, color_palette_type='SemanticKITTI', point_size=3)
draw_bird_eye_view(coords)
def test_NuScenesSCN():
from xmuda.data.utils.visualize import draw_points_image_labels, draw_points_image_depth, draw_bird_eye_view
preprocess_dir = '/datasets_local/datasets_mjaritz/nuscenes_preprocess/preprocess'
nuscenes_dir = '/datasets_local/datasets_mjaritz/nuscenes_preprocess'
# split = ('train_singapore',)
# pselab_paths = ('/home/docker_user/workspace/outputs/xmuda/nuscenes/usa_singapore/xmuda/pselab_data/train_singapore.npy',)
split = ('train_night', )
# pselab_paths = ('/home/docker_user/workspace/outputs/xmuda/nuscenes/day_night/xmuda/pselab_data/train_night.npy',)
dataset = NuScenesSCN(
split=split,
preprocess_dir=preprocess_dir,
nuscenes_dir=nuscenes_dir,
# pselab_paths=pselab_paths,
merge_classes=True,
use_image=True,
noisy_rot=0.1,
flip_x=0.5,
rot_z=2 * np.pi,
transl=True,
fliplr=0.5,
color_jitter=(0.4, 0.4, 0.4))
for i in [10, 20, 30, 40, 50, 60]:
data = dataset[i]
coords = data['coords']
seg_label = data['seg_label']
img = np.moveaxis(data['img'], 0, 2)
img_indices = data['img_indices']
draw_points_image_labels(img,
img_indices,
seg_label,
color_palette_type='NuScenes',
point_size=3)
# pseudo_label_2d = data['pseudo_label_2d']
# draw_points_image_labels(img, img_indices, pseudo_label_2d, color_palette_type='NuScenes', point_size=3)
draw_bird_eye_view(coords)
print('Number of points:', len(coords))