roofAI — 5: A semseg for AIRS: 2

Next: segmentation_models.pytorch/examples/cars segmentation (camvid).ipynb

copy the notebook to nbs : notebooks/camvid.ipynb

validate camvid.ipynb on a smaller dataset,

profile = validation,quick,full

profile = "quick"  # full, quick, validation

epoch_count = {
    "full": 40,
    "quick": 10,
    "validation": 4,
}[profile]

data_count = {
    "full": -1,
    "quick": 20,
    "validation": 4,
}[profile]

validation

took a few seconds.

quick

took a few minutes.

@git push roofAI - "camvid.ipynb validation 🚀"

explore GPU,

@instance help

@instance \
  from_image \
  p2.xlarge \
  - \
 image=abcli-gpu

safety when incorrect info,

@instance \
  from_image \
  p2.xlarge \
  - \
 image=abcli-g4dn

will validate on Mac, then transfer to py, then aws batch through docker.

✅