v0.3.0

• Extension to 3D
• Inputs can have multiple channels (independent denoising, but only one channel segmentation)
• Generate patches from list of images with different sizes