Improve readme

Author: StephanAkkerman

README.md

@@ -12,13 +12,13 @@
 
 ## Introduction
 
-Chart-extractor provides tools for extracting important information from financial charts, utilizing my [chart-info-detector](https://github.com/StephanAkkerman/chart-info-detector) model and OCR tools.
+Chart-extractor provides tools for extracting important information from financial charts, utilizing the [chart-info-detector](https://github.com/StephanAkkerman/chart-info-detector) model and OCR tools. It detects the chart title and the small price "pill" in screenshots, runs OCR on those regions, and applies lightweight parsing heuristics to convert raw OCR text into structured fields such as symbol, exchange, timeframe, price, and trading session. The project is designed primarily for TradingView-style screenshots but also works with similar chart widgets. It supports RapidOCR (ONNX runtime) for fast local inference and can fall back to Tesseract when needed. Use the library programmatically, the small CLI for quick JSON output, or the example script for experiments.
 
 ## Example
 Given an Tradingview chart image like the one below, it will extract the information such as price, symbol, exchange, and timeframe.
 
 <details closed>
-<summary>Input Image Example</summary>
+<summary>Input Image Example 📊</summary>
 <img src="img/chart.png" alt="Input Chart">
 </details>
 

chart_extractor/chart_extractor.py

@@ -1,25 +1,19 @@
-# chart_extractor.py
-from __future__ import annotations
-
 import re
 from dataclasses import dataclass
 from pathlib import Path
 
 import cv2
 import numpy as np
 from huggingface_hub import hf_hub_download
+from rapidocr_onnxruntime import RapidOCR
 from ultralytics import YOLO
 
-# ---------- Config ----------
-# Class IDs (must match your training config)
 CLS_SYMBOL_TITLE = 0
 CLS_LAST_PRICE_PILL = 1
 
-# Hugging Face model (adjust if you renamed the repo or path)
 HF_MODEL_REPO = "StephanAkkerman/chart-info-detector"
-HF_MODEL_FILE = "weights/best.pt"  # path inside the model repo
+HF_MODEL_FILE = "weights/best.pt"
 
-# OCR engine (lazy-loaded): RapidOCR preferred, Tesseract fallback
 _OCR = None
 _OCR_KIND = None  # "rapid" | "tesseract"
 
@@ -55,7 +49,6 @@ def _download_weights_if_needed(
     return hf_hub_download(repo_id=repo_id, filename=filename, repo_type="model")
 
 
-# --- NEW helpers: classify OCR text, smart swap ---
 def _looks_like_price(text: str) -> bool:
     if not text:
         return False
@@ -91,26 +84,10 @@ def _ensure_ocr():
     global _OCR, _OCR_KIND
     if _OCR is not None:
         return _OCR
-    try:
-        from rapidocr_onnxruntime import RapidOCR
-
-        _OCR = RapidOCR()
-        _OCR_KIND = "rapid"
-        return _OCR
-    except Exception:
-        try:
-            import pytesseract  # requires system tesseract (Windows installer / apt-get on Linux)
 
-            _OCR = pytesseract
-            _OCR_KIND = "tesseract"
-            return _OCR
-        except Exception as e2:
-            raise RuntimeError(
-                "No OCR engine available. Install one of:\n"
-                "  pip install rapidocr-onnxruntime onnxruntime\n"
-                "or\n"
-                "  sudo apt-get install tesseract-ocr && pip install pytesseract"
-            ) from e2
+    _OCR = RapidOCR()
+    _OCR_KIND = "rapid"
+    return _OCR
 
 
 def _read_image(img: str | Path | np.ndarray) -> np.ndarray:
@@ -175,7 +152,6 @@ def _ocr_text(im_bgr: np.ndarray) -> str:
 )
 
 
-# --- REPLACE your title parser with this TradingView-oriented version ---
 def _parse_title(text: str) -> tuple[str | None, str | None, str | None]:
     """
     Parse (name, exchange, timeframe) from TradingView-style titles like:
@@ -223,7 +199,6 @@ def _parse_title(text: str) -> tuple[str | None, str | None, str | None]:
     return name, exchange, timeframe
 
 
-# --- TIGHTER price parser (only from pill text) ---
 def _parse_pill(text: str) -> tuple[float | None, str | None]:
     """
     Parse (price, session) from pill text; avoid 'S&P 500' false matches.
@@ -258,7 +233,6 @@ def _parse_pill(text: str) -> tuple[float | None, str | None]:
     return price, ("regular" if sess is None else sess)
 
 
-# ---------- Core pipeline ----------
 class ChartExtractor:
     """
     Detects chart widgets (YOLO) and extracts info via OCR.
@@ -419,7 +393,6 @@ def analyze(
         return result
 
 
-# ---------- Quick CLI test ----------
 if __name__ == "__main__":
     import json
     import sys