Fully Convoluntional Networks for Segmentaion Segmentation

  • Talk: url
  • pdf: url

  • pixelwise prediction

  • supervised pretraining

From classifier to dense FCN

Reinterprete classification nets as fully convolutional and fine-tuning from their learned representations. Fig 7 Fig 8

Fig 10 Fig 11 Fig 12

Combining what and where.

Fig 13 Fig 14

results matching ""

    No results matching ""