Skip to content

Use the key ideas from Transcoda: Synthetic datasets and better data augmentations #107

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
liebharc merged 8 commits into from
Jul 13, 2026
Merged

Use the key ideas from Transcoda: Synthetic datasets and better data augmentations #107

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
23c9006
Used the key ideas from Transcoda: Synthetic datasets and better data…
liebharc Jun 26, 2026
dbfd6d6
Results form a short training run
liebharc Jul 3, 2026
9258673
Merge branch 'main' into feature/synthetic_datasets
liebharc Jul 8, 2026
f7bef4e
'More' scheduled sampling but with a slower decay
liebharc Jul 8, 2026
a288863
Filtering PDMX samples with weird time signatures like 4/444
liebharc Jul 9, 2026
02e87ce
Fixed test
liebharc Jul 9, 2026
b6fd208
Added pdmx and musetrainer to training
liebharc Jul 9, 2026
3a570d9
Updated Training.md
liebharc Jul 13, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
14 changes: 13 additions & 1 deletion Training.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -77,6 +77,19 @@ This validation provides a **more representative indication of overall system pe

Implementation: `rate_validation_result.py`

## Run 426 - 27 epochs

Commit: b6fd20809a8dcaf10dfd39a4ca4f64c6f056e644
Day: 13 July 2026
Transformer Smoke Test: 5%
System Level: Total: 5.7 diffs, SER: 4.2%
System Level after adding https://github.com/liebharc/homr/issues/110: Total: 5.5 diffs, SER: 4.1%
Polish scores: OMR-NED 24.30%

Training with lieder+grandstaff+primux+pdmx+musetrainer datasets.

Note that https://github.com/liebharc/homr/issues/110 will from now on remain in the system level datset to avoid regressions.

## Run 414

Commit: 79aec9b6b66de2281972c9d4f9c606f3f84c9cd1
Expand All @@ -86,7 +99,6 @@ System Level: Total: 5.9 diffs, SER: 4.9%

Clean the Lieder dataset to introduce more training data


## Run 407 - at epoch 6

Commit: fabab4b0f8480be20d13686edf5f91d913ab00fd
Expand Down
3 changes: 2 additions & 1 deletion homr/main.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2,6 +2,7 @@
import glob
import os
import sys
import xml.etree.ElementTree as ET
from concurrent.futures import Future
from dataclasses import dataclass
from enum import Enum
Expand Down Expand Up @@ -215,7 +216,7 @@ def process_image(

eprint("Writing XML", result_staffs)
xml = generate_xml(xml_generator_args, result_staffs, title)
xml.write(xml_file)
ET.ElementTree(xml).write(xml_file, encoding="unicode", xml_declaration=True)

eprint("Finished parsing " + str(len(result_staffs)) + " staves")
teaser_file = replace_extension(image_path, "_teaser.png")
Expand Down
Loading
Loading