FlowNet

Optical Flow

• 2-Channel optical flow
  • $dx,dy$
• Two consecutive frames
  • 6-channel tensor

Skip Connections

• Further through the network information is condensed
  • Less high frequency information
• Link encoder layers to UpConv layers
  • Append activation maps from encoder to decoder

Encode

UpConv

Training

• Synthetic rendered objects
• Real background images