Add new point cloud augmentation - Random Translation

[phuccuongngo99]

What

The current repo doesn't support Random Translation augmentation methods.

This PR adds Random Translation methods here:

Complex-YOLOv4-Pytorch/src/data_process/transformation.py

Lines 376 to 403 in ffb67cd

	class Random_Translate(object):
	def __init__(
	self,
	translate_range_x=(-0.1, 0.1),
	translate_range_y=(-0.1, 0.1),
	p=0.5
	):
	#Translate ranges from [-1, 1]
	#With 1 being the full length of BEV in x/y direction
	self.x_range = translate_range_x
	self.y_range = translate_range_y
	self.p = p
	def __call__(self, lidar, labels):
	if np.random.random() <= self.p:
	factor_x = np.random.uniform(self.x_range[0], self.x_range[1])
	factor_y = np.random.uniform(self.y_range[0], self.y_range[1])
	factor_x *= (cnf.boundary["maxX"] - cnf.boundary["minX"])
	factor_y *= (cnf.boundary["maxY"] - cnf.boundary["minY"])
	lidar[:, 0] += factor_x
	lidar[:, 1] += factor_y
	labels[:, 0] += factor_x
	labels[:, 1] += factor_y
	return lidar, labels

Why

It's reasonable since self-driving car requires a fixed view (the car has to be centered bottom of the image)

However, other applications of 3D point clouds object detection may have different settings such as industrial factories and warehouses like what I tried to do. We may need to augment using translation to increase the diversity of our limited dataset.

Testing

We will make changes to train_lidar_transforms here:

Complex-YOLOv4-Pytorch/src/data_process/kitti_dataloader.py

Lines 22 to 28 in a712f52

	def create_train_dataloader(configs):
	"""Create dataloader for training"""
	train_lidar_transforms = OneOf([
	Random_Rotation(limit_angle=20., p=1.0),
	Random_Scaling(scaling_range=(0.95, 1.05), p=1.0)
	], p=0.66)

After changing, run

cd src/data_process
python kitti_dataloader.py --show-train-data --output-width 608

1) No Augmentation

2) Random_Rotate (Existing augmentation)

train_lidar_transforms = OneOf([
    Random_Rotation(limit_angle=30., p=1.0),
], p=1.0)

3) Random_Scaling (Existing augmentation)

train_lidar_transforms = OneOf([
    Random_Scaling(scaling_range=(0.50, 1.50), p=1.0),
], p=1.0)

You can see that's it's zoomed out

4) Random_Translate (Added in this PR)

train_lidar_transforms = OneOf([
    Random_Translate(translate_range_x=(-0.5, 0.5),
                    translate_range_y=(-0.5, 0.5),
                    p=1.0)
], p=1.0)

You can see the image is translated, the view shifts upwards, no longer starting from the car