​
In this cookbook we demonstrate how FutureAGI SDK can be used to evaluate images:

  • Alignment with Descriptions: Ensuring that generated images accurately represent the textual descriptions provided.
  • Meeting Specific Requirements: Evaluating images against predefined artistic or thematic standards.
  • Evaluating Image Replacements: Assessing the quality and fit of new images replacing previous versions.

​
Installing FutureAGI

pip install ai-evaluation
pip install pillow

​
Making Necessary Imports

import json
import requests

from IPython.display import Image, display

from fi.evals import Evaluator
from fi.evals import Deterministic, ImageInputOutput, ImageInstruction
from fi.testcases import MLLMTestCase

​
Loading and Visualising Data

path = '/content/data.json'
# Open and load the JSON file
with open(path, 'r') as file:
    datapoints = json.load(file)

# Sample Datapoint

datapoint = {
    'id': 'masked_id',
    'image_url': './images/output_8_0.png',
    'output_image_url': './images/output_26_0.png',
    'prompt': 'an asian man, closeup, on new york city street',
    'type': 'T2I',
    'category': 'Ethnicity',
    'question': 'Does the image follow the Ethnicity mentioned in the prompt?'
}

# Sample Image

response = requests.get(datapoint['image_url'])

# Display the image in the notebook
if response.status_code == 200:
    display(Image(response.content))
else:
    print("Failed to fetch the image.")
Output: Sample Image

​
Initializing the FutureAGI Evaluator Class and Deterministic Eval

from getpass import getpass
from fi.evals import Evaluator

fi_api_key = getpass("Enter your FI API Key: ")
fi_secret_key = getpass("Enter your FI Secret Key: ")

evaluator = Evaluator(
    fi_api_key=fi_api_key,
    fi_secret_key=fi_secret_key,
    fi_base_url="https://api.futureagi.com"
)

print("Evaluator client initialized successfully!")

​
Evaluating Alignment with Descriptions

image_eval = ImageInstruction(
    config={
        "criteria": """
        Evaluate the image based on:
        1. Accuracy of object representation
        2. Setting accuracy
        3. Image quality and realism
        """
    }
)

class ImageEvalTestCase(MLLMTestCase):
    input: str
    image_url: str

test_case_img_eval = ImageEvalTestCase(
    input=datapoint['prompt'],
    image_url=datapoint['image_url']
)

import textwrap
batch_result = evaluator.evaluate([image_eval], [test_case_img_eval])
wrapped_text = textwrap.fill(batch_result.eval_results[0].reason, width=80)

print(wrapped_text)
Output: 
The image accurately represents an Asian man and a New York City street, but the anime style affects realism and image quality.

​
Evaluating Subjective Requirements

deterministic_eval = Deterministic(config={
    "multi_choice": False,
    "choices": ["Yes", "No"],
    "rule_prompt": "Prompt : {{input_key2}}, Image : {{input_key3}}. Given the prompt and the corresponding image, answer the Question : {{input_key1}}. Focus only on the {{input_key4}}",
    "input": {
        "input_key1": "question",
        "input_key2": "prompt",
        "input_key3": "image_url",
        "input_key4": "category"
    }
})

class DeterministicTestCase(MLLMTestCase):
    question: str
    prompt: str
    image_url: str
    category: str

test_case = DeterministicTestCase(
    question=datapoint['question'],
    prompt=datapoint['prompt'],
    image_url=datapoint['image_url'],
    category=datapoint['category']
)

batch_result = evaluator.evaluate([deterministic_eval], [test_case])

batch_result.eval_results[0].metrics[0].value
Output: 
['Yes']

print(textwrap.fill(batch_result.eval_results[0].reason, width=80))
Output: 
The image depicts an animated character with traits commonly associated with Asian ethnicity.

​
Evaluating Changes Based on Text Instructions

image_input_output_eval = ImageInputOutput(config={
    "criteria": """
    Evaluate the output image based on:
        1. Adherence to input instruction
        2. Preservation of key elements from input image
        3. Quality of color modification
        4. Image quality and realism
    """
})

class ImageInputOutputTestCase(MLLMTestCase):
    input: str
    input_image_url: str
    output_image_url: str

response = requests.get(datapoint['output_image_url'])

# Display the image in the notebook
if response.status_code == 200:
    display(Image(response.content))
else:
    print("Failed to fetch the image.")
Output: Modified Image 
test_case_image_input_output = ImageInputOutputTestCase(
    input='Replace the man with a man of African ethnicity',
    input_image_url=datapoint['image_url'],
    output_image_url=datapoint['output_image_url']
)

batch_result = evaluator.evaluate([image_input_output_eval], [test_case_image_input_output])
print(textwrap.fill(batch_result.eval_results[0].reason, width=80))
Output: 
The output image accurately replaces the man with one of African ethnicity while preserving all key elements, maintaining high image quality and realism.