jacobmarks / active-learning-plugin Goto Github PK

6.0 6.0 0.0 37 KB

Label your dataset with active learning in FiftyOne!

Python 100.00%

active-learning annotation computer-vision data-cu fiftyone labeling plugin python

active-learning-plugin's Introduction

👋 Hi there!

I'm Jacob, a Senior Machine Learning Engineer & Researcher at Voxel51 , where we're revolutionizing AI and computer vision with our powerful open-source toolset, FiftyOne.

🎓 Education

Ph.D. in Theoretical Physics, Stanford University
B.S. in Intensive Physics, Math & Philosophy, Yale University

📝 Blogging & Writing

I regularly contribute to technical topics on Medium, where I have over 7,500 followers. My writings cover AI, ML, computer vision, data cleaning and curation, and more!

See popular articles

	How I Turned My Company's Docs into a Searchable Database with OpenAI April 25, 2023 \| Towards Data Science In this article, I discuss how I leveraged OpenAI's GPT-3 to turn my company's documentation into a searchable database. This project simplifies the way we access and interact with internal resources, enhancing productivity.
	How I Turned ChatGPT into an SQL-Like Translator for Image and Video Datasets June 08, 2023 \| Towards Data Science In this article, I discuss how I used GPT-3.5 to create a text-to-query translator that allows users to interact with image and video datasets using natural language.
	What I Learned Pushing Prompt Engineering to the Limit June 12, 2023 \| Towards Data Science In this article, I share my experiences and lessons learned from pushing the boundaries of prompt engineering. Using advanced techniques, I explore how to make the most out of language models for various applications.
	AI Telephone — A Battle of Multimodal Models Jun 15, 2023 \| Towards Data Science In this article, I explore the competitive landscape of multimodal AI models by setting up an "AI Telephone" experiment. I discuss the intricacies of various models and how they perform in this unique setup.
	An Ode to my Physics Ph.D. July 18, 2023 \| Towards Data Science In this article, I open up on the journey from physics to machine learning, the challenge of transitioning into industry, and lessons learned along the way!
	How to Build a Semantic Search Engine for Emojis January 09, 2024 \| Towards Data Science In this article, I detail the process of building a custom vector search pipeline utilizing multimodal data, cross-encoders, and reranking!

🤗 Connect!

If you have an idea for an integration, plugin, blog post, or something else you'd like to chat about, feel free to reach out!

Things I Care About

⚛️ Physics
🌎 Climate
📖 Open source | Open science
🫀 Building with purpose

Awesome Open Source Projects

Here is a short list of some open source libraries I love ❤️! I've contributed to some of them, and some I just love using 😎

Data

Models

LLMs

Vector Databases

active-learning-plugin's People

Contributors

Stargazers

Watchers

active-learning-plugin's Issues

How to apply this plugin to object detection

Issue Description

I am keenly interested in your project, specifically the functionality provided by the plugin. I would like to inquire about how to apply this plugin in the context of object detection tasks. While I have reviewed the project documentation, I couldn't find detailed instructions on this specific use case

Error when updating the active learner

Hello,

I followed the steps in the github where I have a voxel51 dataset and no labels. I didn't use the zero-shot annotation because the images I have it won't perform well so I manually annotated around 2K images and my hope was to use this to speed up the annotation process using active learning. I revised the predictions that the active learner made and corrected many that where wrong. But when trying to update the learner it gave me the following error:

Error occurred during operator execution
Traceback (most recent call last):
  File "/opt/miniconda3/envs/dermPipe/lib/python3.9/site-packages/fiftyone/operators/executor.py", line 277, in execute_or_delegate_operator
    result = await do_execute_operator(operator, ctx, exhaust=exhaust)
  File "/opt/miniconda3/envs/dermPipe/lib/python3.9/site-packages/fiftyone/operators/executor.py", line 319, in do_execute_operator
    result = await (
  File "/opt/miniconda3/envs/dermPipe/lib/python3.9/site-packages/fiftyone/core/utils.py", line 2317, in run_sync_task
    return await loop.run_in_executor(_get_sync_task_executor(), func, *args)
  File "/opt/miniconda3/envs/dermPipe/lib/python3.9/concurrent/futures/thread.py", line 58, in run
    result = self.fn(*self.args, **self.kwargs)
  File "/home/tapiasj/fiftyone/__plugins__/@jacobmarks/active_learning/__init__.py", line 481, in execute
    teach_learner(ctx.dataset)
  File "/home/tapiasj/fiftyone/__plugins__/@jacobmarks/active_learning/active_learning.py", line 237, in teach_learner
    learner = cache["learner"]
KeyError: 'learner'

Not entirely sure how to deal with this error and didn't see it in any of the issues in the github. Any help would be appreciated.

Requirements install error: 'sklearn' deprecated, use 'scikit-learn'

Overview

I am trying to install this plugin and the requirements, and the requirements installation errors out due to 'sklearn' being deprecated.

What would be the best way to resolve this for this package to work? And, could I request an update to this package to avoid the error?

I am user interested in using this package. Thank you!

Desired behavior

Install the plugin and requirements without error, using the 'scikit-learn' package instead of 'sklearn'.

To replicate error

Installed the plugin with fiftyone plugins download https://github.com/jacobmarks/active-learning-plugin
And ran install requirements with fiftyone plugins requirements @jacobmarks/active_learning --install

During requirements installation, I hit this error:

Collecting sklearn
  Downloading sklearn-0.0.post12.tar.gz (2.6 kB)
  Preparing metadata (setup.py) ... error
Traceback (most recent call last):
  File "/usr/local/bin/fiftyone", line 8, in <module>
    sys.exit(main())
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/core/cli.py", line 4267, in main
    args.execute(args)
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/core/cli.py", line 4250, in <lambda>
    parser.set_defaults(execute=lambda args: command.execute(parser, args))
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/core/cli.py", line 3359, in execute
    fop.install_plugin_requirements(name, error_level=error_level)
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/plugins/core.py", line 309, in install_plugin_requirements
    fou.install_requirements(req_path, error_level=error_level)
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/core/utils.py", line 476, in install_requirements
    install_package(req_str, error_level=error_level)
  File "/usr/local/lib/python3.10/dist-packages/fiftyone/core/utils.py", line 377, in install_package
    return etau.install_package(
  File "/usr/local/lib/python3.10/dist-packages/eta/core/utils.py", line 662, in install_package
    handle_error(PackageError(error_msg), error_level)
  File "/usr/local/lib/python3.10/dist-packages/eta/core/utils.py", line 1004, in handle_error
    raise error
eta.core.utils.PackageError: Failed to install package 'sklearn'

  error: subprocess-exited-with-error
  
  × python setup.py egg_info did not run successfully.
  │ exit code: 1
  ╰─> [15 lines of output]
      The 'sklearn' PyPI package is deprecated, use 'scikit-learn'
      rather than 'sklearn' for pip commands.
      
      Here is how to fix this error in the main use cases:
      - use 'pip install scikit-learn' rather than 'pip install sklearn'
      - replace 'sklearn' by 'scikit-learn' in your pip requirements files
        (requirements.txt, setup.py, setup.cfg, Pipfile, etc ...)
      - if the 'sklearn' package is used by one of your dependencies,
        it would be great if you take some time to track which package uses
        'sklearn' instead of 'scikit-learn' and report it to their issue tracker
      - as a last resort, set the environment variable
        SKLEARN_ALLOW_DEPRECATED_SKLEARN_PACKAGE_INSTALL=True to avoid this error
      
      More information is available at
      https://github.com/scikit-learn/sklearn-pypi-package
      [end of output]
  
  note: This error originates from a subprocess, and is likely not a problem with pip.
error: metadata-generation-failed

× Encountered error while generating package metadata.
╰─> See above for output.

note: This is an issue with the package mentioned above, not pip.
hint: See above for details.

[notice] A new release of pip is available: 23.2.1 -> 23.3.1
[notice] To update, run: python -m pip install --upgrade pip

How to use the model predictions to select the samples

@jacobmarks - I became your fan after looking at your work. Great work.

I am trying to use the active learning pipeline for a classification problem. Based on the blog, I understood that currently 51 active learning plugin can provide a fecility to select various ML based algorithms to train, predict and select the samples based on the embeddings.

But I need to use the ground truth labels and custom model prections (class names and confidence) to select the samples. Can you tell me how it can be done with the existing plugin.

	How I Turned My Company's Docs into a Searchable Database with OpenAI April 25, 2023 \| Towards Data Science In this article, I discuss how I leveraged OpenAI's GPT-3 to turn my company's documentation into a searchable database. This project simplifies the way we access and interact with internal resources, enhancing productivity.
	How I Turned ChatGPT into an SQL-Like Translator for Image and Video Datasets June 08, 2023 \| Towards Data Science In this article, I discuss how I used GPT-3.5 to create a text-to-query translator that allows users to interact with image and video datasets using natural language.
	What I Learned Pushing Prompt Engineering to the Limit June 12, 2023 \| Towards Data Science In this article, I share my experiences and lessons learned from pushing the boundaries of prompt engineering. Using advanced techniques, I explore how to make the most out of language models for various applications.
	AI Telephone — A Battle of Multimodal Models Jun 15, 2023 \| Towards Data Science In this article, I explore the competitive landscape of multimodal AI models by setting up an "AI Telephone" experiment. I discuss the intricacies of various models and how they perform in this unique setup.
	An Ode to my Physics Ph.D. July 18, 2023 \| Towards Data Science In this article, I open up on the journey from physics to machine learning, the challenge of transitioning into industry, and lessons learned along the way!
	How to Build a Semantic Search Engine for Emojis January 09, 2024 \| Towards Data Science In this article, I detail the process of building a custom vector search pipeline utilizing multimodal data, cross-encoders, and reranking!