small_module

Note: The scripts and notebooks in this repository are intended for logic demonstration and cannot be executed directly — most required input files (score images, YOLO staff JSONs, bracket JSONs, GPT staffgroup JSONs) are not included.

Two example CSV outputs are provided for reference:

• a_2_cv_result.csv — CV prediction output from a/
• b_trans_gpt.csv — GPT prediction output from b/

To generate the necessary input files, see github.com/Bobo1111111/OMR_layout.

This repository contains modular scripts and notebooks for orchestral score layout analysis — detecting and labelling staves, systems, staff groups, instrument names, parts, and transposition tones from scanned orchestral score images.

The pipeline combines computer vision (CV) and GPT vision API approaches, with evaluation scripts at each stage.

Folder Overview

Folder	Purpose
a/	CV-based detection of systems and staff groups using vertical barlines
b/	Transform GPT-generated JSON predictions into CSV; evaluate against ground truth
bracket/	Detect bracket/brace symbols in score images to infer staff groupings
c/	Exploratory notebook reconciling CV (a) and GPT (b) results step-by-step, producing c_1.csv through c_6.csv
d/	Reorganization of c/ into a cleaner production pipeline; uses bracket data and GPT-5.1 vision to produce the final d.csv
g/	Independent experiments testing different GPT-5.1 vision prompt formats for staff group correction; g_5_enpower_auto_func.py is used by d/

Data Flow

Score Images + YOLO Staff JSON
        │
        ├─► a/  (CV barline detection) ─────────────► a_2_cv_result.csv
        │                                                      │
        ├─► GPT staffgroup JSON                                │
        │       b/ (transform + evaluate) ──► b_trans_gpt.csv │
        │                                           │          │
        ├─► bracket/ (bracket detection) ──► bracket JSON files│
        │                   │                       │          │
        │           g/ (prompt experiments)         │          │
        │               │                           │          │
        │    g_5_enpower_auto_func.py ◄─────────────┘          │
        │               │                                      │
        └───────────────┴──────────────────────────────────────┘
                        │
                  c/ (exploratory: c_1.csv → c_6.csv)
                  d/ (production:  d.csv)
                        │
                  g_eva.py (evaluate final output)