NVIDIA Computer Vision
Frequently Asked Questions
Learn more about the world of computer vision solutions for developers.
Cameras are becoming increasingly sophisticated and ubiquitous, resulting in an exponential increase in their application across various industries. These include agriculture, autonomous driving, consumer electronics, gaming, healthcare, manufacturing, and retail services to name a few. In all these applications, computer vision (CV) is the technology that enables the cameras and vision systems to perceive, process, analyze, and interpret information in images and videos. For example, the face-tracking feature in your smart phone's camera app is a simple computer vision application. CV involves identifying and extracting information from images or frames of a video to perform actions like detecting and understanding objects, and identifying interesting regions in an image.
A major challenge with traditional computer vision approaches has been that they typically involve a human expert to design custom algorithms for identifying and extracting features of interest in an image or video. As the number of images to work with increases and their characteristics change, traditional CV techniques become increasingly complex, cumbersome, and time-consuming to create and optimize.
More recently, deep learning—a field within AI—has had a great impact on approach to solving CV-based problems in numerous real-world applications. Unlike traditional CV algorithms, deep neural networks have enabled automatic feature extraction using the training and ground truth data and developing CV models without relying heavily on human experts. Deep learning- or AI-based CV applications have also been shown to excel in performance in terms of both accuracy and speed when compared to the traditional algorithms. Additionally, with the emergence of Graphics Processing Units (GPUs) to develop AI-based CV applications, developers have been successful in resolving challenges related to training and compute power. In fact, AI-based computer vision is what helps make self-driving cars a reality.
Learn more about NVIDIA’s work in Deep Learning and Artificial Intelligence.
A variety of different computer vision-related application frameworks are available from NVIDIA through Software Development Kits (SDKs). Application frameworks can also be accessed through NVIDIA’s SDKs. Find the SDK that meets your needs here.
Some of these applications are also available from NVIDIA GPU Cloud (NGC), which enables easy access to AI software, including CV models, to NVIDIA developers.
Optimized AI Software For Developer Needs
- Built and maintained by experts
- Covers popular applications and use cases
- Supercharged with the latest features
- Continuously updated software
- Instant access to the latest features and highest performance
- Winner of the MLPerf competition
Tested Across Platforms
- Supports multi-GPU and multi-node systems
- Passes stringent security scans
- Powers Cloud | Edge | On-prem | workstations
- Access to NVIDIA AI experts
- Faster time-to-solution
- Minimize system downtimes
If you’re new to computer vision, object detection, image classification, and image segmentation tasks are a great way to start. NVIDIA provides pretrained AI-based CV models for these and many other tasks that you can include in your development project. These pretrained models are free to access through NVIDIA's Transfer Learning Toolkit (TLT) on NGC.
Pretrained AI models are ideal for incorporating computer vision in your projects, especially because creating one from scratch takes a lot of training data, time, and computing power. You can also customize these pretrained models by tweaking the model layers, optimizing the model for speed and storage (pruning), or incorporating your own data through transfer learning—the process of transferring or retraining learned features from one application to another. In CV, it’s common to retrain and finetune a model to do a different task. NVIDIA provides Transfer Learning Toolkit as a powerful and customizable tool to retrain models. Best of all, zero to little coding and programming experience is needed to get started.
In addition to pretrained models and Transfer Learning Toolkit, the DeepStream SDK from NVIDIA makes it quick and easy to get started with deploying CV models to an embedded GPU or cloud.
When it comes to AI-based computer vision models, GPUs are the preferred development hardware for both training and deployment. The type of GPU hardware required depends on the specific application/solution area. Learn more about NVIDIA's specific developer hardware platforms and their requirement for different solutions here.
NVIDIA has a tremendous, passionate developer community. Learn more about their work and the benefits of the NVIDIA developer program here.
We support you with the solutions and the training you need.
Sign up for the latest, tailored newsletter updates from NVIDIA here.
Follow and track us on:
Find out more about who we are by reviewing the Corporate Social Responsibility that’s integral to our success and transforming the world.