Batch inspection mode

For normal inspection, the trained deep neural network inspects images one by one through GPU's computation.
If your GPU's computing power and memory size are large enough to process multiple images at once, you can utilize more GPU resources to reduce inspect time. As explained, the method of inspecting multiple images at once using more GPU resources is refered to Batch inspection mode in this document. Only Classification and Segmentation modules currently support this function.

There are only two things you need to do to use Batch inspection mode.

1. Pass the number of images to inspect at once to Memory-allocation functions as batchSize.
   • If batchSize is too large to allocated memory on GPU, Memory-allocation functions returns eErrorType::OUT_OF_GPU_MEMORY.
   • Recommend to use batchSize as power of 2. ex) 2, 4, 8 ...
2. Use dedicated Inspection functions for Batch inspection mode.
   • For these functions, the multiple images are passed at once.
   • If the number of passed images is larger than the batchSize set in step 1, the For loop is executed in units of batchSize until all images are inspected.
   • Segmentation
     • SR_ApplySegTestFromMultiple8bitCharGrayArrays
     • SR_ApplySegTestFromMultiple8bitCharColorArrays
     • SR_ApplySegTestFromMultipleUnsignedIntColorArrays
   • Classification
     • SR_ApplyClsTestFromMultiple8bitCharGrayArrays
     • SR_ApplyClsTestFromMultiple8bitCharColorArrays
     • SR_ApplyClsTestFromMultipleUnsignedIntColorArrays
     • SR_ApplyClsTestFromSegmentedObjects