From 66a8285421e96f804edde863926202e7ecad09fd Mon Sep 17 00:00:00 2001 From: Francisco Ingham <24279597+fpingham@users.noreply.github.com> Date: Sun, 18 Feb 2024 13:36:43 -0300 Subject: [PATCH] doc: image extraction (#346) Co-authored-by: Jason Liu --- docs/examples/extract_slides.md | 113 +++++++++++++++ examples/vision/slides.py | 250 ++++++++++++++++++++++++++++++++ mkdocs.yml | 1 + 3 files changed, 364 insertions(+) create mode 100644 docs/examples/extract_slides.md create mode 100644 examples/vision/slides.py diff --git a/docs/examples/extract_slides.md b/docs/examples/extract_slides.md new file mode 100644 index 0000000..4127399 --- /dev/null +++ b/docs/examples/extract_slides.md @@ -0,0 +1,113 @@ +# Data extraction from slides + +In this guide, we demonstrate how to extract data from slides. + +!!! tips "Motivation" + + When we want to translate key information from slides into structured data, simply isolating the text and running extraction might not be enough. Sometimes the important data is in the images on the slides, so we should consider including them in our extraction pipeline. + +## Defining the necessary Data Structures + +Let's say we want to extract the competitors from various presentations and categorize them according to their respective industries. + +Our data model will have `Industry` which will be a list of `Competitor`'s for a specific industry, and `Competition` which will aggregate the competitors for all the industries. + +```python +from openai import OpenAI +from pydantic import BaseModel, Field +from typing import Optional, List + +class Competitor(BaseModel): + name: str + features: Optional[List[str]] + + +# Define models +class Industry(BaseModel): + """ + Represents competitors from a specific industry extracted from an image using AI. + """ + + name: str = Field( + description="The name of the industry" + ) + competitor_list: List[Competitor] = Field( + description="A list of competitors for this industry" + ) + +class Competition(BaseModel): + """ + This class serves as a structured representation of + competitors and their qualities. + """ + + industry_list: List[IndustryCompetition] = Field( + description="A list of industries and their competitors" + ) +``` + +## Competitors extraction + +To extract competitors from slides we will define a function which will read images from urls and extract the relevant information from them. + +```python +import instructor +from openai import OpenAI + +# Apply the patch to the OpenAI client +# enables response_model keyword +client = instructor.patch( + OpenAI(), mode=instructor.Mode.MD_JSON +) + +# Define functions +def read_images(image_urls: List[str]) -> Competition: + """ + Given a list of image URLs, identify the competitors in the images. + """ + return client.chat.completions.create( + model="gpt-4-vision-preview", + response_model=Competition, + max_tokens=2048, + temperature=0, + messages=[ + { + "role": "user", + "content": [ + { + "type": "text", + "text": "Identify competitors and generate key features for each competitor.", + }, + *[ + {"type": "image_url", "image_url": {"url": url}} + for url in image_urls + ], + ], + } + ], + ) +``` + +## Execution + +Finally, we will run the previous function with a few sample slides to see the data extractor in action. + +As we can see, our model extracted the relevant information for each competitor regardless of how this information was formatted in the original presentations. + +```python +url = [ + 'https://miro.medium.com/v2/resize:fit:1276/0*h1Rsv-fZWzQUyOkt', + 'https://earlygame.vc/wp-content/uploads/2020/06/startup-pitch-deck-5.jpg' + ] +model = read_images(url) +print(model.model_json_dump(indent=2)) +``` + industry_list=[ + + Industry(name='Accommodation and Hospitality', competitor_list=[Competitor(name='CouchSurfing', features=['Affordable', 'Online Transaction']), Competitor(name='Craigslist', features=['Affordable', 'Offline Transaction']), Competitor(name='BedandBreakfast.com', features=['Affordable', 'Offline Transaction']), Competitor(name='AirBed&Breakfast', features=['Affordable', 'Online Transaction']), Competitor(name='Hostels.com', features=['Affordable', 'Online Transaction']), Competitor(name='VRBO', features=['Expensive', 'Offline Transaction']), Competitor(name='Rentahome', features=['Expensive', 'Online Transaction']), Competitor(name='Orbitz', features=['Expensive', 'Online Transaction']), Competitor(name='Hotels.com', features=['Expensive', 'Online Transaction'])]), + + Industry(name='Wine E-commerce', competitor_list=[Competitor(name='WineSimple', features=['Ecommerce Retailers', 'True Personalized Selections', 'Brand Name Wine', 'No Inventory Cost', 'Target Mass Market']), Competitor(name='NakedWines', features=['Ecommerce Retailers', 'Target Mass Market']), Competitor(name='Club W', features=['Ecommerce Retailers', 'Brand Name Wine', 'Target Mass Market']), Competitor(name='Tasting Room', features=['Ecommerce Retailers', 'True Personalized Selections', 'Brand Name Wine']), Competitor(name='Drync', features=['Ecommerce Retailers', 'True Personalized Selections', 'No Inventory Cost']), Competitor(name='Hello Vino', features=['Ecommerce Retailers', 'Brand Name Wine', 'Target Mass Market'])]) + + ] +``` +``` diff --git a/examples/vision/slides.py b/examples/vision/slides.py new file mode 100644 index 0000000..4a342c9 --- /dev/null +++ b/examples/vision/slides.py @@ -0,0 +1,250 @@ +import json +import logging +import sys +from typing import List, Optional + +from dotenv import find_dotenv, load_dotenv +from openai import OpenAI +from pydantic import BaseModel, Field +from rich import print as rprint + +import instructor + +load_dotenv(find_dotenv()) + +IMAGE_FILE = "image-file.txt" # file with all the images to be processed + +# Add logger +logging.basicConfig() +logger = logging.getLogger("app") +logger.setLevel("INFO") + +class Competitor(BaseModel): + name: str + features: Optional[List[str]] + + +# Define models +class Industry(BaseModel): + """ + Represents competitors from a specific industry extracted from an image using AI. + """ + + name: str = Field( + description="The name of the industry" + ) + competitor_list: List[Competitor] = Field( + description="A list of competitors for this industry" + ) + +class Competition(BaseModel): + """ + Represents competitors extracted from an image using AI. + + This class serves as a structured representation of + competitors and their qualities. + """ + + industry_list: List[Industry] = Field( + description="A list of industries and their competitors" + ) + +# Define clients +client_image = instructor.patch( + OpenAI(), mode=instructor.Mode.MD_JSON +) + +# Define functions +def read_images(image_urls: List[str]) -> Competition: + """ + Given a list of image URLs, identify the competitors in the images. + """ + + logger.info(f"Identifying competitors in images... {len(image_urls)} images") + + return client_image.chat.completions.create( + model="gpt-4-vision-preview", + response_model=Competition, + max_tokens=2048, + temperature=0, + messages=[ + { + "role": "user", + "content": [ + { + "type": "text", + "text": "Identify competitors and generate key features for each competitor.", + }, + *[ + {"type": "image_url", "image_url": {"url": url}} + for url in image_urls + ], + ], + } + ], + ) + + + +def process_and_identify_competitors(): + """ + Main function to process the image list file and identify competitors. + """ + + logger.info("Starting app...") + + try: + with open(IMAGE_FILE, "r") as file: + logger.info(f"Reading images from file: {IMAGE_FILE}") + image_list = file.read().splitlines() + logger.info(f"{len(image_list)} images read from file: {IMAGE_FILE}") + except Exception as e: + logger.error(f"Error reading images from file: {IMAGE_FILE}") + logger.error(e) + sys.exit(1) + + competitors = read_images(image_list) + + rprint(f"[green]{len(competitors.industry_list)} industries identified:[/green]") + for industry in competitors.industry_list: + rprint(f"[green]{industry.name}[/green]") + rprint(f"[blue]Features: {industry.competitor_list}[/blue]") + + logger.info("Writing results to file...") + + with open("results.json", "w") as f: + json.dump( + { + "competitors": competitors.model_dump(), + }, + f, + indent=4, + ) + +if __name__ == "__main__": + process_and_identify_competitors() + +""" +Example output: +{ + "competitors": { + "industry_list": [ + { + "name": "Accommodation and Hospitality", + "competitor_list": [ + { + "name": "craigslist", + "features": [ + "Transactions Offline", + "Inexpensive" + ] + }, + { + "name": "couchsurfing", + "features": [ + "Transactions Offline", + "Inexpensive" + ] + }, + { + "name": "BedandBreakfast.com", + "features": [ + "Transactions Offline", + "Inexpensive" + ] + }, + { + "name": "airbnb", + "features": [ + "Transactions Online", + "Inexpensive" + ] + }, + { + "name": "HOSTELS.com", + "features": [ + "Transactions Online", + "Inexpensive" + ] + }, + { + "name": "VRBO", + "features": [ + "Transactions Offline", + "Costly" + ] + }, + { + "name": "Rentahome", + "features": [ + "Transactions Online", + "Costly" + ] + }, + { + "name": "Orbitz", + "features": [ + "Transactions Online", + "Costly" + ] + }, + { + "name": "Hotels.com", + "features": [ + "Transactions Online", + "Costly" + ] + } + ] + }, + { + "name": "E-commerce Wine Retailers", + "competitor_list": [ + { + "name": "winesimple", + "features": [ + "Ecommerce Retailers", + "True Personalized Selections", + "Brand Name Wine", + "No Inventory Cost", + "Target Mass Market" + ] + }, + { + "name": "nakedwines.com", + "features": [ + "Ecommerce Retailers", + "Target Mass Market" + ] + }, + { + "name": "Club W", + "features": [ + "Ecommerce Retailers", + "Brand Name Wine", + "Target Mass Market" + ] + }, + { + "name": "Tasting Room", + "features": [ + "Ecommerce Retailers", + "True Personalized Selections", + "Brand Name Wine" + ] + }, + { + "name": "hellovino", + "features": [ + "Ecommerce Retailers", + "True Personalized Selections", + "No Inventory Cost", + "Target Mass Market" + ] + } + ] + } + ] + } +} +""" diff --git a/mkdocs.yml b/mkdocs.yml index 7aed607..a2710c5 100644 --- a/mkdocs.yml +++ b/mkdocs.yml @@ -151,6 +151,7 @@ nav: - Batch Classification (User Defined): 'examples/batch_classification.md' - LLM Self Critique: 'examples/self_critique.md' - Extracting Tables with GPT-V: 'examples/extracting_tables.md' + - Extracting From Slides with GPT-V: 'examples/extract_slides.md' - Content Moderation: 'examples/moderation.md' - Citing Sources (RAG): 'examples/exact_citations.md' - Extracting Knowledge Graphs: 'examples/knowledge_graph.md'