InviolusOCR: tiny Turkish line OCR for microcontrollers

A compact convolutional CTC recognizer for one cropped line of printed Turkish text: 560,954 parameters and a 598 KB full-int8 TensorFlow Lite file that keeps almost all of the float model's accuracy.

Türkçe özet: Mikrodenetleyiciler için tek satır Türkçe metin okuyan, 561 bin parametreli, int8 TFLite dosyası 598 KB olan küçük bir OCR modeli. int8 hâli, ayrılmış test satırlarının %78,4'ünü ve gerçek sahne metni (TS-TR) test kırpıntılarının %58,2'sini harfi harfine doğru okuyor. Ticari olmayan kullanım içindir (CC BY-NC-SA 4.0).

This replaces inviolus/erdem-ocr-tr-tinyml (2026-10-09), whose int8 file lost most of its accuracy.

Newer version on GitHub (2026-11-01): a continuation of this model, D, trained with 47,613 extra printed-document lines (donut 200k, CC BY-SA 4.0), reads 46.5% of the donut test rows exactly as int8 (this model: 9.3%), with the same size and 1 to 2 points less on the sets below. The files here are still the 2026-10-31 model. Scores: InviolusOCR results.

Files

Model

• Input: one horizontal text line, grayscale, resized to 160×32, scaled to [-1, 1] (pixel / 127.5 - 1).
• Architecture: 4 conv + batch-norm + pool blocks, a dense projection, then four residual dilated 1-D convolutions (dilation 1, 2, 4, 1) instead of an LSTM, and a dense softmax. Only ops that quantize cleanly to int8.
• Output: 40 timesteps × 90 classes. Class 0 is padding, 1..88 are the characters of alphabet.json in order, 89 is the CTC blank.
• Character set: digits, ASCII letters, ç ğ ı İ ö ş ü Ç Ğ Ö Ş Ü, space and .,:;!?-_/()'%. Up to 24 characters per line.
• Not included: text detection. The line must already be cropped.

Usage

import json
import numpy as np
import tensorflow as tf

alphabet = json.load(open("alphabet.json", encoding="utf-8"))["alphabet"]
BLANK = len(alphabet) + 1

def preprocess(path):
    image = tf.io.decode_image(tf.io.read_file(path), channels=1, expand_animations=False)
    image = tf.image.resize(image, [32, 160])
    return (tf.cast(image, tf.float32) / 127.5 - 1.0).numpy()

def decode(ids):
    text, previous = [], None
    for i in ids:
        if i != previous and 0 < i < BLANK:
            text.append(alphabet[i - 1])
        previous = i
    return "".join(text)

# int8 TFLite (the file meant for devices)
interpreter = tf.lite.Interpreter(model_path="inviolusocr_int8.tflite")
interpreter.allocate_tensors()
inp, out = interpreter.get_input_details()[0], interpreter.get_output_details()[0]
scale, zero = inp["quantization"]
x = np.clip(np.round(preprocess("line.png") / scale + zero), -128, 127).astype(np.int8)
interpreter.set_tensor(inp["index"], x[np.newaxis])
interpreter.invoke()
print(decode(interpreter.get_tensor(out["index"])[0].argmax(-1)))

Results

Exact-line accuracy (the whole line correct), all models scored on the same rows: 12,093 held-out lines (synthetic, Zenodo, TS-TR, esengul3, STS-TR) and the 2,078-crop official TS-TR test split (real scene text, never trained on).

Character error rate of the int8 file: 9.1% held-out, 18.1% TS-TR.

The 2026-10-25 model is not published because it was trained with a dataset that states no license.

Limitations

• Printed text only; no handwriting, curved text or full pages.
• Real scene text is still the weak spot (58% exact).
• Licence plates are not read (0% exact on 250 plate crops in a first, unchecked test).
• Text detection and line cropping must be done before this model runs.

Training data

Trained for 200 minutes on CPU, starting from the 2026-10-09 weights, on about 66,000 crops:

• Synthetic Turkish lines rendered from a 50k-word Turkish word list.
• Turkish OCR Text Image Dataset, Zenodo, CC BY 4.0.
• Turkish Scene Text Recognition (TS-TR), Serdar Yıldız, CC BY-NC 4.0.
• Synthetic Turkish Scene Text (STS-TR), Serdar Yıldız, CC BY-NC 4.0.
• Turkish Word OCR, esengul3, CC BY-SA 4.0.

License

Because of the non-commercial (TS-TR, STS-TR) and share-alike (esengul3) terms of the training data, this model is released under CC BY-NC-SA 4.0: non-commercial use only, credit this model and the datasets above, and share adaptations under the same license.

Training code (MIT): ErdemWilkinson/InviolusOCR.