liuliu/ccv

★ 7,216C++Audience · developerComplexity · 4/5LicenseSetup · moderate

Mindmap

mindmap
  root((ccv))
    Detection
      Face detection
      Pedestrian detection
      Text detection
    Tracking
      Object tracking
      Video frames
    Portability
      Raspberry Pi
      Mobile devices
      Desktop Linux
    Design
      No dependencies
      Built-in cache

mindmap root((ccv)) Detection Face detection Pedestrian detection Text detection Tracking Object tracking Video frames Portability Raspberry Pi Mobile devices Desktop Linux Design No dependencies Built-in cache

Click or tap to explore — scroll the page freely

Things people build with this

USE CASE 1

Detect faces in images on a Raspberry Pi or mobile device using CCV's dependency-free C implementation.

USE CASE 2

Track a moving object across video frames in a robotics or surveillance application.

USE CASE 3

Detect text regions in images as a preprocessing step for document scanning or OCR.

Tech stack

CC++LinuxmacOSiOSAndroid

Getting it running

Difficulty · moderate Time to first run · 1h+

Core detection algorithms need only a C compiler, the convolutional neural network features require an optional external math library.

Use freely in commercial and open-source software, keep the copyright notice and do not use the project name to endorse derived products.

In plain English

CCV stands for C-based, Cached, Core Computer Vision Library. It is an open-source library written primarily in C that provides computer vision algorithms for detecting and analyzing objects in images and video. The project started in 2010 as a personal effort to build a leaner, easier-to-deploy alternative to the computer vision tools available at the time. Portability is one of the library's defining traits. CCV runs on Mac, Linux, FreeBSD, Windows, iPhone, iPad, Android, and Raspberry Pi. Most of its code compiles with no additional dependencies beyond a standard C compiler, which is what makes it usable on embedded devices and phones alongside full desktop systems. The more computationally intensive parts, specifically the convolutional neural network code, require an external math library, but that subset is optional. The library includes a collection of algorithms chosen based on real application needs rather than academic novelty. These cover detecting faces in images, classifying what is in a photograph, finding pedestrians and cars, detecting text regions, tracking a moving object across video frames over long periods, and extracting feature points for matching images to each other. The README describes the development philosophy as application-driven, meaning algorithms were added because they were needed for practical work. One design choice that sets CCV apart is a built-in cache for image preprocessing steps. Many vision tasks require the same preparation work on each image: converting color formats, building image pyramids for different scales, and similar operations. CCV handles these transparently behind the scenes so calling code stays simple while avoiding redundant computation. The library also connects to a separate deep learning project at libnnc.org. CCV is distributed under the BSD 3-clause license, which is permissive and allows use in commercial software.

Copy-paste prompts

Prompt 1

Using the CCV library in C, how do I detect faces in a JPEG file and print the bounding box coordinates of each detected face?

Prompt 2

How do I compile CCV on a Raspberry Pi for pedestrian detection without any GPU or external math library?

Prompt 3

Show me how to use CCV's built-in image cache to avoid recomputing color conversions when running multiple detection algorithms on the same image.

Open on GitHub → Explain another repo

← liuliu on gitmyhub — every repo by this author, as a profile.

Verify against the repo before relying on details.