Computer Vision Challenge 🏆

Overview

This is a collection of foundational projects for anyone diving into computer vision.

Explore some of computer vision core concepts and hands-on projects through challenges.

Challenges are organized into levels:

• Level 0 - Zero/beginner: Getting Started with Basics
• Level 1 - Apprentice/intermediate: Hands-on Computer Vision with Deep Learning
• Level 2 - Hero: Large Vision Models (LVMs) from Image Generation, Inpainting, & More
• Level 3 - Advanced: Video Models Benchmarking (ongoing)
• Level 4 - Expert: Finetuning of VLMs (Vision Language Models) & LVMs (ongoing)
• Level 5 - Master: Multimodality (ongoing)

Important

In L1 and L2, we primarily leverage pre-trained models to ensure accessibility for everyone. This also allows us to explore a wider range of vision recognition tasks using different types of models while focusing on the model's performance and outcome.

Basic Computer Vision Pipeline

graph LR
    A[Image Acquisition] ==> B[Image Processing]
    B ==> C[Feature Extraction]
    C ==> D[Output, Interpretation & Analysis]

    style A fill:#EEE,stroke:#333,stroke-width:4px
    style B fill:#F88,stroke:#333,stroke-width:4px
    style C fill:#4F4,stroke:#333,stroke-width:4px
    style D fill:#33F,stroke:#333,stroke-width:4px

Loading

Requirements

To install the dependency packages using either conda or pip:

Using conda:

1. create a new conda environment

conda create --name cv-challenge

2. Activate the newly created environment:

source activate cv-challenge  # For bash/zsh
conda activate cv-challenge  # For conda prompt/powershell

3. Install dependencies from the requirements.txt file:

conda install --channel conda-forge --file requirements.txt

Using pip:

1. Install dependencies from the requirements.txt file:

pip install -r requirements.txt

Hands-on Computer Vision Challenges!

Level 0 - Zero: Getting Started with Basics 💪

	Project	Description	Notebooks
[1]	Getting Started with Images	Load an image, display it, and apply basic transformations.
[2]	Basic Image Manipulation	Modify pixels, resizing, Flipping, Cropping, image annotations
[3]	Image Filtering & Restoration	Enhance or manipulate image features using filtering techniques.
[4]	Image Enhancement	Enhance using arithmetic & bitwise operations
[5]	Image Segmentation (Traditional)	segment images into regions or pixels that belong to different classes or categories
[6]	Feature Extraction & Alignment	Learn how to extract features from images using descriptors based on the nature of the features
[7]	Optical Character Recognition (OCR)	Learn how to recognize text in images or documents using libraries such as Tesseract, Pytesseract, or EasyOCR

Level 1 - Apprentice: Hands-on Computer Vision with Deep Learning 🔥

	Project	Description	Notebooks
[1]	MNIST Handwritten Digit Recognition	Train a simple neural network to classify handwritten digits from the MNIST dataset.
[2]	CIFAR-10 Image Classification	Utilize convolutional neural networks (CNNs) to classify images of different types of objects from the CIFAR-10 dataset.
[3]	Object Detection with YOLOv5	Implement YOLOv5, a real-time object detection algorithm, to detect objects in images and videos.
[4]	Semantic Segmentation with DeepLabv3+	Utilize DeepLabv3+, a semantic segmentation model, to segment images into different semantic categories.
[5]	Facial Recognition with OpenFace	Explore facial recognition using OpenFace, a facial recognition library, to identify individuals in images.
[6]	Object Tracking	Follow the movement of objects in a video sequence.
[7]	Human Pose Estimation	Estimate the pose of a person in an image or a video using OpenCV and a pre-trained model.

Level 2 - Hero: Large Vision Models (LVMs) from Image Generation, Inpainting, & More ⚡

	Project	Description	Notebooks
[1]	Creative Image Generation with GANs	Generate novel images of different styles using GANs.
[2]	Text-to-Image Synthesis with LLMs and Diffusion Models	Create realistic and creative images from text descriptions using LLMs and diffusion models.
[3]	AI-Powered Image Restoration and Enhancement	Restore and enhance images using AI methods.
[4]	Style Transfer with GANs and Image Processing	Transfer the artistic style of one image to another.
[5]	AI-Driven Image Captioning and Storytelling	Generate comprehensive and creative captions and stories from images using LLMs.
[6]	AI-Assisted Image Editing and Manipulation	Automate image editing and manipulation tasks using AI.
[7]	AI Image Recognition Benchmarks with SOTA Vision Models	Benchmark SOTA Vision Models on a variety of image recognition tasks, including image classification, object detection, ...

Level 3 - Advanced: Video Models Benchmarking

	Project	Description	Notebooks
[1]	Video Generation & Captioning	Create realistic video content from text, generate descriptive text or subtitles for video content using AI models.
[2]	Facial Emotion Recognition	Automatically generate descriptive text or subtitles for video content using AI.
[3]	Motion Analysis	Analyze the motion and movement of objects in a video sequence techniques: tracking, optical flow, video detection, etc.
[4]	Video Segmentation	Divide video frames into meaningful segments or regions for analysis and processing.
[5]	Video Style Transfer	Apply artistic styles from one video or image to another video, transforming its visual appearance.
[6]	Video Restoration & Enhancement	Restore and enhance videos using AI methods.
[7]	Video Models Benchmarking	Benchmark SOTA Video Models on a variety of video recognition tasks, including video classification, object detection, etc.

Usage

Most projects are written in Jupyter notebooks, you can run the directly using jupyter notebook/lab or Colab.

For projects with a main.py file, run the command below:

python main.py

Roadmap & Upcoming Features

Roadmap:

    flowchart BT
        A(Level 0: Zero) --> B(Level 1: Intermediate)
        A --> C(Level 2: Hero)
        A --> D(Level 3: Advanced)
        A --> E(Level 4: Expert)
        A --> F(Level 5: Master)
        
        style A fill:#fff,stroke:#333,stroke-width:2px
        style B fill:#88f,stroke:#333,stroke-width:2px
        style C fill:#8f8,stroke:#333,stroke-width:2px
        style D fill:#bbb,stroke:#f66,stroke-width:2px,color:#fff,stroke-dasharray: 5 5
        style E fill:#bbb,stroke:#f66,stroke-width:2px,color:#fff,stroke-dasharray: 5 5
        style F fill:#bbb,stroke:#f66,stroke-width:2px,color:#fff,stroke-dasharray: 5 5

Loading

New levels:

• L3 - Advanced: Video Models Benchmarking
• L4 - Expert: Finetuning of VLMs (Vision Language Models) & LVMs
• L5 - Master: Multimodality

Upcoming Features:

Feature	Description	Status
Code Refactoring	Enhance code readability by cleaning, documenting, and integrating Gradio demos.	To-Do
New Learning Levels	Introduce advanced levels: L3 - Video Models Benchmarking, L4 - Finetuning of VLMs (Vision Language Models) & LVMs, and L5 - Multimodality	To-Do
Wiki Update	Document the new learning levels in the project Wiki.	To-Do
Multilingual Support	Translate the README.md file into multiple languages (French, Spanish, etc.).	To-Do
Edge Device Deployment	Explore code translation for deployment on edge devices using C++ or Rust.	To-Do
Performance Enhancements	Investigate options to improve performance, including adding new datasets and supporting additional computer vision tasks.	To-Do
Machine Learning Framework Integration	Integrate the project with popular machine learning frameworks.	To-Do

Contributing

We warmly welcome your contributions! Whether you're a seasoned developer or just starting out in Computer Vision, you can help us improve the project and make it more valuable to everyone.

How to contribute:

• Fork this repository and clone it to your local machine.
• Create a new branch with a descriptive name for your contribution.
• Add your code and files to the branch and commit your changes.
• Push your branch to your forked repository and create a pull request to the main repository.
• Wait for your pull request to be reviewed and merged.

Sponsor this Project

Another way to get involved is by sponsoring the project.

Your support will help:

• Provide computational resources (This is a GPU Poor Project!!!) to explore new frontiers in computer vision by training larger and more complex model
• Keep the project up to date with the latest computer vision advancements
• Create more detailed tutorials for users at all skill levels

LICENSE

This project is licensed under the MIT LICENSE.

Star History

"Vision is a picture of the future that produces passion." - Bill Hybels