---
dataset_info:
  features:
  - name: image
    dtype: image
  - name: label
    dtype: image
  - name: name
    dtype: string
  splits:
  - name: train
    num_bytes: 1258400018.614
    num_examples: 1282
  - name: validation
    num_bytes: 229176381.0
    num_examples: 226
  download_size: 1456945904
  dataset_size: 1487576399.614
configs:
- config_name: default
  data_files:
  - split: train
    path: data/train-*
  - split: validation
    path: data/validation-*
---

# elwha-segmentation-v1

Contains 1k+ manually selected images from `elwha-segmentation-all` which are higher quality (mask must generally follow the river path, no huge misclassified patches).