vhad/SoftGroup
1
0
Fork 0
mirror of https://github.com/botastic/SoftGroup.git synced 2025-10-16 11:45:42 +00:00
Code Issues Actions Packages Projects Releases Wiki Activity

add train aug_prob remove val aug

Browse Source
This commit is contained in:
Thang Vu 2022-04-11 01:17:44 +00:00
parent 3332a478a7
commit 40572e2834
1 changed files with 14 additions and 11 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

25
softgroup/data/custom.py
View File

@ -85,13 +85,13 @@ class CustomDataset(Dataset):
pt_offset_label = pt_mean - xyz pt_offset_label = pt_mean - xyz
return instance_num, instance_pointnum, instance_cls, pt_offset_label return instance_num, instance_pointnum, instance_cls, pt_offset_label
def dataAugment(self, xyz, jitter=False, flip=False, rot=False): def dataAugment(self, xyz, jitter=False, flip=False, rot=False, prob=0.9):
m = np.eye(3) m = np.eye(3)
if jitter: if jitter and np.random.rand() < prob:
m += np.random.randn(3, 3) * 0.1 m += np.random.randn(3, 3) * 0.1
if flip: if flip and np.random.rand() < prob:
m[0][0] *= np.random.randint(0, 2) * 2 - 1 # flip x randomly m[0][0] *= np.random.randint(0, 2) * 2 - 1 # flip x randomly
if rot: if rot and np.random.rand() < prob:
theta = np.random.rand() * 2 * math.pi theta = np.random.rand() * 2 * math.pi
m = np.matmul(m, [[math.cos(theta), math.sin(theta), 0], m = np.matmul(m, [[math.cos(theta), math.sin(theta), 0],
[-math.sin(theta), math.cos(theta), 0], [0, 0, 1]]) # rotation [-math.sin(theta), math.cos(theta), 0], [0, 0, 1]]) # rotation
@ -122,12 +122,15 @@ class CustomDataset(Dataset):
j += 1 j += 1
return instance_label return instance_label
def transform_train(self, xyz, rgb, semantic_label, instance_label): def transform_train(self, xyz, rgb, semantic_label, instance_label, aug_prob=0.9):
xyz_middle = self.dataAugment(xyz, True, True, True) xyz_middle = self.dataAugment(xyz, True, True, True, aug_prob)
xyz = xyz_middle * self.voxel_cfg.scale xyz = xyz_middle * self.voxel_cfg.scale
xyz = self.elastic(xyz, 6 * self.voxel_cfg.scale // 50, 40 * self.voxel_cfg.scale / 50) if np.random.rand() < aug_prob:
xyz = self.elastic(xyz, 20 * self.voxel_cfg.scale // 50, 160 * self.voxel_cfg.scale / 50) xyz = self.elastic(xyz, 6 * self.voxel_cfg.scale // 50, 40 * self.voxel_cfg.scale / 50)
xyz -= xyz.min(0) xyz = self.elastic(xyz, 20 * self.voxel_cfg.scale // 50,
160 * self.voxel_cfg.scale / 50)
xyz_middle = xyz / self.voxel_cfg.scale
xyz = xyz - xyz.min(0)
max_tries = 5 max_tries = 5
while (max_tries > 0): while (max_tries > 0):
xyz_offset, valid_idxs = self.crop(xyz) xyz_offset, valid_idxs = self.crop(xyz)
@ -145,11 +148,11 @@ class CustomDataset(Dataset):
return xyz, xyz_middle, rgb, semantic_label, instance_label return xyz, xyz_middle, rgb, semantic_label, instance_label
def transform_test(self, xyz, rgb, semantic_label, instance_label): def transform_test(self, xyz, rgb, semantic_label, instance_label):
xyz_middle = self.dataAugment(xyz, False, True, True) xyz_middle = self.dataAugment(xyz, False, False, False)
xyz = xyz_middle * self.voxel_cfg.scale xyz = xyz_middle * self.voxel_cfg.scale
xyz -= xyz.min(0) xyz -= xyz.min(0)
valid_idxs = np.ones(xyz.shape[0], dtype=bool) valid_idxs = np.ones(xyz.shape[0], dtype=bool)
instance_label = self.getCroppedInstLabel(instance_label, valid_idxs) # TODO remove this instance_label = self.getCroppedInstLabel(instance_label, valid_idxs)
return xyz, xyz_middle, rgb, semantic_label, instance_label return xyz, xyz_middle, rgb, semantic_label, instance_label
def __getitem__(self, index): def __getitem__(self, index):