Fully Convolutional Network for semantic segmentation

A FCN based on the original paper using PyTorch and trained on the CamVid dataset.

Result

• Pixel accuracy of 91.69%
• Mean IoU of 0.3755

Inference

Prediction completely labeled.

Prediction partially labeled.

Inference on video