|
| 1 | +# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. ========= |
| 2 | +# Licensed under the Apache License, Version 2.0 (the "License"); |
| 3 | +# you may not use this file except in compliance with the License. |
| 4 | +# You may obtain a copy of the License at |
| 5 | +# |
| 6 | +# http://www.apache.org/licenses/LICENSE-2.0 |
| 7 | +# |
| 8 | +# Unless required by applicable law or agreed to in writing, software |
| 9 | +# distributed under the License is distributed on an "AS IS" BASIS, |
| 10 | +# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. |
| 11 | +# See the License for the specific language governing permissions and |
| 12 | +# limitations under the License. |
| 13 | +# ========= Copyright 2023-2024 @ CAMEL-AI.org. All Rights Reserved. ========= |
| 14 | + |
| 15 | +from io import BytesIO |
| 16 | +from typing import List, Optional |
| 17 | +from urllib.parse import urlparse |
| 18 | + |
| 19 | +import requests |
| 20 | +from PIL import Image |
| 21 | + |
| 22 | +from camel.logger import get_logger |
| 23 | +from camel.messages import BaseMessage |
| 24 | +from camel.models import BaseModelBackend, ModelFactory |
| 25 | +from camel.toolkits import FunctionTool |
| 26 | +from camel.toolkits.base import BaseToolkit |
| 27 | +from camel.types import ModelPlatformType, ModelType |
| 28 | + |
| 29 | +logger = get_logger(__name__) |
| 30 | + |
| 31 | + |
| 32 | +class ImageAnalysisToolkit(BaseToolkit): |
| 33 | + r"""A toolkit for comprehensive image analysis and understanding. |
| 34 | + The toolkit uses vision-capable language models to perform these tasks. |
| 35 | + """ |
| 36 | + |
| 37 | + def __init__(self, model: Optional[BaseModelBackend] = None): |
| 38 | + r"""Initialize the ImageAnalysisToolkit. |
| 39 | +
|
| 40 | + Args: |
| 41 | + model (Optional[BaseModelBackend]): The model backend to use for |
| 42 | + image analysis tasks. This model should support processing |
| 43 | + images for tasks like image description and visual question |
| 44 | + answering. If None, a default model will be created using |
| 45 | + ModelFactory. (default: :obj:`None`) |
| 46 | + """ |
| 47 | + if model: |
| 48 | + self.model = model |
| 49 | + else: |
| 50 | + self.model = ModelFactory.create( |
| 51 | + model_platform=ModelPlatformType.DEFAULT, |
| 52 | + model_type=ModelType.DEFAULT, |
| 53 | + ) |
| 54 | + |
| 55 | + def image_to_text( |
| 56 | + self, image_path: str, sys_prompt: Optional[str] = None |
| 57 | + ) -> str: |
| 58 | + r"""Generates textual description of an image with optional custom |
| 59 | + prompt. |
| 60 | +
|
| 61 | + Args: |
| 62 | + image_path (str): Local path or URL to an image file. |
| 63 | + sys_prompt (Optional[str]): Custom system prompt for the analysis. |
| 64 | + (default: :obj:`None`) |
| 65 | +
|
| 66 | + Returns: |
| 67 | + str: Natural language description of the image. |
| 68 | + """ |
| 69 | + default_content = '''You are an image analysis expert. Provide a |
| 70 | + detailed description including text if present.''' |
| 71 | + |
| 72 | + system_msg = BaseMessage.make_assistant_message( |
| 73 | + role_name="Senior Computer Vision Analyst", |
| 74 | + content=sys_prompt if sys_prompt else default_content, |
| 75 | + ) |
| 76 | + |
| 77 | + return self._analyze_image( |
| 78 | + image_path=image_path, |
| 79 | + prompt="Please describe the contents of this image.", |
| 80 | + system_message=system_msg, |
| 81 | + ) |
| 82 | + |
| 83 | + def ask_question_about_image( |
| 84 | + self, image_path: str, question: str, sys_prompt: Optional[str] = None |
| 85 | + ) -> str: |
| 86 | + r"""Answers image questions with optional custom instructions. |
| 87 | +
|
| 88 | + Args: |
| 89 | + image_path (str): Local path or URL to an image file. |
| 90 | + question (str): Query about the image content. |
| 91 | + sys_prompt (Optional[str]): Custom system prompt for the analysis. |
| 92 | + (default: :obj:`None`) |
| 93 | +
|
| 94 | + Returns: |
| 95 | + str: Detailed answer based on visual understanding |
| 96 | + """ |
| 97 | + default_content = """Answer questions about images by: |
| 98 | + 1. Careful visual inspection |
| 99 | + 2. Contextual reasoning |
| 100 | + 3. Text transcription where relevant |
| 101 | + 4. Logical deduction from visual evidence""" |
| 102 | + |
| 103 | + system_msg = BaseMessage.make_assistant_message( |
| 104 | + role_name="Visual QA Specialist", |
| 105 | + content=sys_prompt if sys_prompt else default_content, |
| 106 | + ) |
| 107 | + |
| 108 | + return self._analyze_image( |
| 109 | + image_path=image_path, |
| 110 | + prompt=question, |
| 111 | + system_message=system_msg, |
| 112 | + ) |
| 113 | + |
| 114 | + def _load_image(self, image_path: str) -> Image.Image: |
| 115 | + r"""Loads an image from either local path or URL. |
| 116 | +
|
| 117 | + Args: |
| 118 | + image_path (str): Local path or URL to image. |
| 119 | +
|
| 120 | + Returns: |
| 121 | + Image.Image: Loaded PIL Image object. |
| 122 | +
|
| 123 | + Raises: |
| 124 | + ValueError: For invalid paths/URLs or unreadable images. |
| 125 | + requests.exceptions.RequestException: For URL fetch failures. |
| 126 | + """ |
| 127 | + parsed = urlparse(image_path) |
| 128 | + |
| 129 | + if parsed.scheme in ("http", "https"): |
| 130 | + logger.debug(f"Fetching image from URL: {image_path}") |
| 131 | + try: |
| 132 | + response = requests.get(image_path, timeout=15) |
| 133 | + response.raise_for_status() |
| 134 | + return Image.open(BytesIO(response.content)) |
| 135 | + except requests.exceptions.RequestException as e: |
| 136 | + logger.error(f"URL fetch failed: {e}") |
| 137 | + raise |
| 138 | + else: |
| 139 | + logger.debug(f"Loading local image: {image_path}") |
| 140 | + try: |
| 141 | + with Image.open(image_path) as img: |
| 142 | + # Load immediately to detect errors |
| 143 | + img.load() |
| 144 | + return img.copy() |
| 145 | + except Exception as e: |
| 146 | + logger.error(f"Image loading failed: {e}") |
| 147 | + raise ValueError(f"Invalid image file: {e}") |
| 148 | + |
| 149 | + def _analyze_image( |
| 150 | + self, |
| 151 | + image_path: str, |
| 152 | + prompt: str, |
| 153 | + system_message: BaseMessage, |
| 154 | + ) -> str: |
| 155 | + r"""Core analysis method handling image loading and processing. |
| 156 | +
|
| 157 | + Args: |
| 158 | + image_path (str): Image location. |
| 159 | + prompt (str): Analysis query/instructions. |
| 160 | + system_message (BaseMessage): Custom system prompt for the |
| 161 | + analysis. |
| 162 | +
|
| 163 | + Returns: |
| 164 | + str: Analysis result or error message. |
| 165 | + """ |
| 166 | + try: |
| 167 | + image = self._load_image(image_path) |
| 168 | + logger.info(f"Analyzing image: {image_path}") |
| 169 | + |
| 170 | + from camel.agents.chat_agent import ChatAgent |
| 171 | + |
| 172 | + agent = ChatAgent( |
| 173 | + system_message=system_message, |
| 174 | + model=self.model, |
| 175 | + ) |
| 176 | + |
| 177 | + user_msg = BaseMessage.make_user_message( |
| 178 | + role_name="User", |
| 179 | + content=prompt, |
| 180 | + image_list=[image], |
| 181 | + ) |
| 182 | + |
| 183 | + response = agent.step(user_msg) |
| 184 | + agent.reset() |
| 185 | + return response.msgs[0].content |
| 186 | + |
| 187 | + except (ValueError, requests.exceptions.RequestException) as e: |
| 188 | + logger.error(f"Image handling error: {e}") |
| 189 | + return f"Image error: {e!s}" |
| 190 | + except Exception as e: |
| 191 | + logger.error(f"Unexpected error: {e}") |
| 192 | + return f"Analysis failed: {e!s}" |
| 193 | + |
| 194 | + def get_tools(self) -> List[FunctionTool]: |
| 195 | + r"""Returns a list of FunctionTool objects representing the functions |
| 196 | + in the toolkit. |
| 197 | +
|
| 198 | + Returns: |
| 199 | + List[FunctionTool]: A list of FunctionTool objects representing the |
| 200 | + functions in the toolkit. |
| 201 | + """ |
| 202 | + return [ |
| 203 | + FunctionTool(self.image_to_text), |
| 204 | + FunctionTool(self.ask_question_about_image), |
| 205 | + ] |
0 commit comments