Wiki
AI
Neural Networks
Convolutional
FCN

FCN

Fully Convolutional Network

Convolutional and up-convolutional layers with 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

fcn-uses fcn-arch

Training

  • Rarely from scratch
  • Pre-trained weights
  • Replace final layers
    • FC layers
    • White-noise initialised
  • Add UpConv layer(s)
    • Fine-tune train
    • Freeze others
    • Annotated GT images
  • Can use summed per-pixel log loss

Evaluation

fcn-eval

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