### Install Ppllocr Library

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Installs the Ppllocr library using pip. This is the first step before using any of its functionalities.

```bash
pip install ppllocr
```

--------------------------------

### Initialize Ppllocr OCR Class

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Demonstrates how to initialize the OCR class, either with default bundled model and CPU inference, or with a custom model path and optional GPU acceleration. It also shows how to check the installed library version.

```python
from ppllocr import OCR

# Default initialization (uses bundled model, CPU inference)
ocr = OCR()

# Custom model path with GPU support
ocr_custom = OCR(model_path="/path/to/custom_model.onnx", use_gpu=True)

# Check version
import ppllocr
print(ppllocr.__version__)  # Output: "2.2"
```

--------------------------------

### Authenticate and Solve CAPTCHA with Requests

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Demonstrates how to use the requests library to handle authenticated sessions, fetch a CAPTCHA image, solve it using ppllocr's OCR, and submit login credentials.

```python
import requests
from pplocr import OCR

ocr = OCR()

# With session for authenticated requests
session = requests.Session()
session.headers.update({
    "User-Agent": "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36"
})

# Login flow with CAPTCHA
captcha_img = session.get("https://example.com/captcha").content
captcha_text = ocr.classification(captcha_img)

login_data = {
    "username": "user",
    "password": "pass",
    "captcha": captcha_text
}
response = session.post("https://example.com/login", data=login_data)

```

--------------------------------

### Web Scraping Integration with Ppllocr

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Demonstrates how to integrate Ppllocr with web scraping by fetching CAPTCHA images directly from URLs using the `requests` library and processing the image content as bytes. This allows for solving CAPTCHAs in real-time from web sources.

```python
import requests
from ppllocr import OCR

ocr = OCR()

# Fetch and recognize CAPTCHA from URL
def solve_captcha(url, headers=None):
    response = requests.get(url, headers=headers)
    if response.status_code == 200:
        return ocr.classification(response.content)
    return None

# Example: Luogu CAPTCHA
captcha_url = "https://www.luogu.com.cn/lg4/captcha"
result = solve_captcha(captcha_url)
print(f"CAPTCHA solved: {result}")
```

--------------------------------

### Detailed Recognition with Bounding Boxes using classification_box()

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Returns both the recognized text and detailed information for each detected character, including confidence scores and bounding box coordinates. This is useful for character-level validation or visualization. It also supports custom thresholds.

```python
from ppllocr import OCR

ocr = OCR()

with open("captcha.jpg", "rb") as f:
    img_bytes = f.read()

# Returns tuple: (text, details_list)
text, details = ocr.classification_box(img_bytes)

print(f"Full text: {text}")  # Output: "2a3B"

# Details structure:
# [
#   {'char': '2', 'conf': 0.98, 'box': [10.5, 5.0, 30.2, 45.1]},
#   {'char': 'a', 'conf': 0.95, 'box': [35.0, 8.0, 55.0, 48.0]},
#   {'char': '3', 'conf': 0.92, 'box': [60.0, 6.0, 80.5, 46.0]},
#   {'char': 'B', 'conf': 0.97, 'box': [85.0, 7.0, 105.0, 47.0]}
# ]

for char_info in details:
    char = char_info['char']
    conf = char_info['conf']
    box = char_info['box']  # [x1, y1, x2, y2]
    print(f"Character: {char}, Confidence: {conf:.2f}, Box: {box}")

# Filter low-confidence results
high_conf_chars = [d for d in details if d['conf'] > 0.9]
print(f"High confidence characters: {''.join([d['char'] for d in high_conf_chars])}")

# With custom thresholds
text, details = ocr.classification_box(img_bytes, conf=0.5, iou=0.4)
```

--------------------------------

### Basic Text Recognition with classification()

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Performs CAPTCHA recognition and returns the recognized text as a string. This method accepts input from file paths, bytes, or numpy arrays. It also allows for custom confidence and IoU thresholds.

```python
from ppllocr import OCR

ocr = OCR()

# From file path
text = ocr.classification("captcha.jpg")
print(f"Result: {text}")  # Output: "2a3B"

# From bytes (common in web scraping)
with open("captcha.jpg", "rb") as f:
    img_bytes = f.read()
text = ocr.classification(img_bytes)
print(f"Result: {text}")  # Output: "X7kP"

# From numpy array
import cv2
img_array = cv2.imread("captcha.jpg")
text = ocr.classification(img_array)
print(f"Result: {text}")  # Output: "9mNq"

# With custom confidence and IoU thresholds
# Higher conf = fewer false positives, lower recall
# conf: confidence threshold (default 0.25)
# iou: NMS overlap threshold (default 0.45)
text = ocr.classification(img_bytes, conf=0.6, iou=0.45)
print(f"Result: {text}")
```

--------------------------------

### Legacy Interface predict()

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

The predict() method is an alias for classification_box(), maintained for backward compatibility. It provides the same functionality of returning text and detailed character information with bounding boxes.

```python
from ppllocr import OCR

ocr = OCR()

with open("captcha.jpg", "rb") as f:
    img_bytes = f.read()

# Equivalent to classification_box()
text, details = ocr.predict(img_bytes, conf=0.25, iou=0.45)
print(f"Result: {text}")
print(f"Details: {details}")
```

--------------------------------

### Ppllocr Supported Character Set

Source: https://context7.com/gitpetyr/ppllocr/llms.txt

Defines the full character set supported by ppllocr v2.2, including alphanumeric characters, symbols, and characters for arithmetic expressions.

```python
# Full character set supported by ppllocr v2.2
CHARACTERS = [
    '#', '%', '*', '+', '-', '/',
    '0', '1', '2', '3', '4', '5', '6', '7', '8', '9',
    '=', '?', '@',
    'A', 'B', 'C', 'D', 'E', 'F', 'G', 'H', 'I', 'J', 'K', 'L', 'M',
    'N', 'O', 'P', 'Q', 'R', 'S', 'T', 'U', 'V', 'W', 'X', 'Y', 'Z',
    'a', 'b', 'c', 'd', 'e', 'f', 'g', 'h', 'i', 'j', 'k', 'l', 'm',
    'n', 'o', 'p', 'q', 'r', 's', 't', 'u', 'v', 'w', 'x', 'y', 'z'
]

# Supports arithmetic expressions (v2.2 enhancement)
# Characters: + - ( ) for math CAPTCHAs like "3+5=" or "8-2="

```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.