stem/AI/Neural Networks/CNN/FCN

Fully Convolutional Network

Convolutional Layer and UpConv with Activation Functions#ReLu but no others (pooling)

• All some sort of Encoder-Decoder

Contractive → UpConv

Image Segmentation

• For visual output
  • Previously image \rightarrow vector
• Additional layers to up-sample representation to an image
  • Up-convolutional
  • De-convolutional

!

!

Training

• Rarely from scratch
• Pre-trained weights
• Replace final layers
  • MLP layers
  • White-noise initialised
• Add upconv layer(s)
  • Fine-tune train
  • Freeze others
  • Annotated GT images
• Can use summed per-pixel log Deep Learning#Loss Function

Evaluation

!

• SDS
  • Classical method
  • 52% mAP
• FCN
  • 62% mAP
• Intersection over Union
  • IOU
  • Jaccard
  • Averaged over all images
  • J(A,B)=\frac{|A\cap B|}{|A\cup B|}