Requirements differ based on how you plan to use the application, as the number of images, resolution, modules used can affect significantly the amount of GPU memory required.
If you use modules for Images Preprocessing, Measurement or Code, in most cases, you can safely use PC without GPU.
Hardware Requirements
NVIDIA GPU with at least 8GB (CUDA capability is 4.0 or higher)
Minimum 16 GB of RAM
Minimum recommended NVIDIA GeForce® RTX 2080 Ti
Supported embedded hardware: Jetson Modules (JetPack 4.4 & 4.5)
Software Requirements
Windows 10
Linux – Ubuntu 18.04 and latest
For price-sensitive use-cases it is possible to use a less powerful NVIDIA GPU card, However, this can result in longer processing time or possible cause the PC to run out of memory.
If you are require to perform deep-learning training for very large datasets with high resolution and/or running several projects simultaneously for the same machine, it may require a stronger GPU with more memory than we recommend for typical cases. Therefore, this should be considered when planning your project.