A barcode or QR scanner tested on a clean printed sample works reliably. The same code, crumpled, poorly lit, or photographed at an angle on a real receipt, fails far more often — and that's the condition it actually needs to work in.
What actually helps
- Preprocessing: adaptive thresholding and deskewing before detection, rather than feeding the raw photo directly to a decoder.
- Trying multiple crop regions and rotations if the first detection attempt fails, rather than giving up after one pass.
- Falling back to a larger, more robust detection library rather than a minimal one if your first choice struggles on real-world samples.
Testing honestly
Build your test set from real photos taken the way users will actually take them — hand-held, imperfect lighting — not from clean scanned samples, which will always overstate real-world performance.
See camera-quality problems that quietly break a vision model.
GenAI engineer and AI product builder with 2+ years shipping production-grade AI systems: RAG pipelines, fine-tuned LLMs, hybrid retrieval and multi-modal apps across vision, speech and OCR, architected end to end from ingestion to deployment. Leads engineering for SCULT INDIA's 14-member team, founded the 500+ member TechVerse Enclave community and has mentored 200+ students. Three hackathon first prizes: Changethon 2025 (IIT Roorkee), Product Genesis at Vividhotsava 2025 (CSJMU Kanpur) and BYTEBATTLE (MeetKats).
Open to GenAI roles, hackathon judging, mentorship sessions and guest talks at colleges. Email pranjulrathour41@gmail.com.
