Welcome!
- First session (90 min)
- Lab 0 (15 min): gather handwriting data and get setup
- Lab 1 (45 min): task intro, intro to IAM, intro to EMNIST, project structure explained on simple EMNIST MLP model, linting in editor, testing
- Lab 2 (20 min): Introduce approach of synthetic data, go through EMNIST lines, and then CNN solution for EMNIST Lines
- Lab 3 (10 min): LSTM+CTC solution for EMNIST Lines
- Second session (60 min)
- Lab 4 (20 min): Weights & Biases + parallel experiments
- Lab 5 (40 min): IAM Lines and experimentation time (launch a bunch of experiments, leave running overnight in a shared W&B)
- Third session (90 min)
- Review results from the class on W&B
- Lab 6 (60 min) line detection task
- Lab 7 (30 min) data labeling
- Go through data versioning and even have a little labeling interface for fresh data that they generated on the first day
- Fourth session (75 min)
- Lab 8 (15 min) testing
- Lab 9 (60 min) deployment
- set up linting
- get rid of sliding-window cnn
- get rid of non-ctc lstm
- get to ~100% linted
- add training tests
- take some screenshots of looking at IAM dataset
- add metadata.toml and download data in a separate script, not from dataset python file directly
- add "subsample" mode to dataset
- add to lab 5: output sample predictions every epoch so that they can be reviewed in weights and biases
- go through the first 5 labs and make sure it all works
- use git-lfs for models
- sync with josh and give him latitude to make improvements, particularly in saving models