Tools

Repository

Line-Level-OCR-using-CNN-BiLSTM-CTC-loss

by Udayps2303

CNN-BiLSTM-CTC model for printed Odia script recognition from line-level images, built with TensorFlow/Keras. Achieves CER < 2% on ~4 lakh line samples.

No licensePython 0

Last updated 12 May 2026 · published 12 May 2026

Cite this

No licenseCite

Tags

cnn-bilstm-ctccrnndeep-learningimage-processingindic-ocrkerasline-segmentationmachine-learningnlpocrodiaoptical-character-recognitionpythontensorflow