2-240725-3n

v1.0.1

- integrated new data QS_7415 (empty); QS_7506, QS_7508, QS_7510 (low density)
- optimized batched inference
- improved model training (architecture, loss, brightness augmentation, normalization, evaluation)
- improved plotting
- improved released model provision
- improved sbatch scripts
- included drafts for identification and removal of hard-to-learn points (rmbad)

2-240725-3

v1.0.1

- integrated new data QS_7415 (empty); QS_7506, QS_7508, QS_7510 (low density)
- optimized batched inference
- improved model training (architecture, loss, brightness augmentation, normalization, evaluation)
- improved plotting
- improved released model provision
- improved sbatch scripts
- included drafts for identification and removal of hard-to-learn points (rmbad)

2-240725-1n

v1.0.1

- integrated new data QS_7415 (empty); QS_7506, QS_7508, QS_7510 (low density)
- optimized batched inference
- improved model training (architecture, loss, brightness augmentation, normalization, evaluation)
- improved plotting
- improved released model provision
- improved sbatch scripts
- included drafts for identification and removal of hard-to-learn points (rmbad)

2-240725-1

Recommended best model version. 1% worse training accuracy than 2-240725-3, but much smaller and probably less overfitted.

2-240628-v002

Re-release of v0.0.2 model with new model API for use with the newest software version.