What are the recommended system requirements for ERDAS IMAGINE's Deep Learning operators?
The IMAGINE Spatial Modeler operators Initialize Object Detection and Detect Objects Using Deep Learning are enabled to use GPU acceleration for operator tasks. NVIDIA’s CUDA enabled GPU is the only graphics adapter supported for GPU acceleration with the IMAGINE Spatial Model operators and IMAGINE’s SAR Feature module. It is recommended that the NVIDIA graphics card has 4 GB or greater video memory available.
If the installed NVIDIA graphics adapter does not meet this memory requirement, GPU acceleration should be disabled in the IMAGINE preferences. IMAGINE Preferences > General > GPU > Use CUDA
These operators also have a CPU hardware dependency for the Advanced Vector Instruction Set (AVX) instruction set extension. The operators will not run without this CPU instruction set. The AVX instruction set is supported on Windows operating systems (Windows 7 SP1 or greater).