This week focused on planning the whole system with the team as we prepared our proposal presentation and continuously defining the technical direction of the software components of the project. Specifically, this week I focused on researching and learning more about how to implement the sheet music scanning mechanism and the OMR system as part of the LumiKeys architecture.
Planning & Team Design Work
- Worked with Chris to design the end-to-end software pipeline and how integrates with the hardware layer with Gene, as well as work on refining proposal for the presentation.
- Helped define major pipeline stages including:
- Image preprocessing
- Staff detection and segmentation
- Musical symbol recognition
- Temporal grouping and quantization
- MusicXML generation
- Discussed module boundaries and data flow to ensure each component can be developed and tested independently and how we can reach performance goals related to accuracy and latency for the scanning system
Initial Technical Prototyping
- Performed exploratory preprocessing experiments on sheet music images using Python, OpenCV, and NumPy, including grayscale normalization, adaptive thresholding, and morphological filtering to evaluate staff and symbol separability
- Analyzed sensitivity of preprocessing results to lighting variation and capture quality
- Assembled a dataset of smartphone-scanned sheet music for controlled evaluation
Schedule
I am currently on schedule as core scanning architecture and requirements have been defined and implementation work has already started.
Deliverables for Next Week
- Begin implementing staff detection and basic symbol recognition
- Continue refining preprocessing techniques
- Start building automated testing framework