Add ImageToDescription processor example

image_to_description.py

@@ -0,0 +1,51 @@
+from typing import Annotated, List
+
+from pydantic import BaseModel
+
+from marker.processors.llm import BaseLLMSimpleBlockProcessor, PromptData, BlockData
+from marker.schema import BlockTypes
+from marker.schema.document import Document
+
+
+class ImageToDescriptionProcessor(BaseLLMSimpleBlockProcessor):
+    """Simple processor that replaces images with generated text."""
+
+    block_types = (
+        BlockTypes.Picture,
+        BlockTypes.Figure,
+    )
+
+    extract_images: Annotated[bool, "Extract images from the document."] = True
+
+    def inference_blocks(self, document: Document) -> List[BlockData]:
+        blocks = super().inference_blocks(document)
+        if self.extract_images:
+            return []
+        return blocks
+
+    def block_prompts(self, document: Document) -> List[PromptData]:
+        prompts = []
+        for block_data in self.inference_blocks(document):
+            block = block_data["block"]
+            prompts.append(
+                {
+                    "prompt": "",
+                    "image": self.extract_image(document, block),
+                    "block": block,
+                    "schema": ImageDescriptionSchema,
+                    "page": block_data["page"],
+                }
+            )
+        return prompts
+
+    def rewrite_block(
+        self, response: dict, prompt_data: PromptData, document: Document
+    ):
+        block = prompt_data["block"]
+        image_name = block.id.to_path()
+        description = f"This is the image {image_name}"
+        block.html = f"<p>{description}</p>"
+
+
+class ImageDescriptionSchema(BaseModel):
+    description: str

image_to_description_script.py

@@ -0,0 +1,44 @@
+#!/usr/bin/env python
+"""
+Minimal script: run Marker with ``ImageToDescriptionProcessor``.
+"""
+
+from pathlib import Path
+
+from image_to_description import ImageToDescriptionProcessor
+from marker.converters.pdf import PdfConverter
+from marker.models import create_model_dict
+from marker.config.parser import ConfigParser
+
+PDF_PATH = Path(
+    "Design Patterns in C# A Hands-on Guide with Real-World Examples - Vaskaran Sarcar.pdf"
+)
+PAGE_RANGE = [31]
+OUTPUT_DIR = Path("test_output")
+
+# Prepare config
+config = {
+    "extract_images": False,  # Must be False for processor to run!
+    "output_dir": str(OUTPUT_DIR),
+    "page_range": ",".join(str(p) for p in PAGE_RANGE),
+    "output_format": "markdown",
+    "use_llm": True,
+    # Add any other LLM config keys needed (e.g., API key)
+    # "gemini_api_key": "your_key_here",
+}
+
+# Create converter with the processor class
+converter = PdfConverter(
+    artifact_dict=create_model_dict(),
+    config=ConfigParser(config).generate_config_dict(),
+    processor_list=[ImageToDescriptionProcessor],
+    # Replace with your actual LLM service if needed
+    llm_service="marker.services.ollama.OllamaService",
+)
+
+# Run conversion and save markdown
+OUTPUT_DIR.mkdir(parents=True, exist_ok=True)
+rendered = converter(str(PDF_PATH))
+final_md_path = OUTPUT_DIR / f"{PDF_PATH.stem}_processed.md"
+final_md_path.write_text(rendered.markdown, encoding="utf-8")
+print(f"✅ Processed markdown saved to: {final_md_path}")