Known Limitations - Known Limitations - 6.3 English - XD056

The release focuses on functionality and some feature may not be fully optimized for performance.

Supported batch size is limited to 1 in vvas_xinfer plug-in.

Supported inference level is limited to 1 in vvas_xinfer plug-in.

Software pre-processing (libvvas_image_process_sw_lib.so) is not supported in this release.

This release has been validated with color video formats and does not support monochrome input.

Running multiple models concurrently is resulting in occasional failure and putting the board in bad state.

Using ctrl-C while application is ongoing may result in abnormal termination.

The image_processing kernel requires the minimum cropping width and height to be 16. If the cropping width or height is less than 16, then cropping is not performed.

If the Image Processing kernel is being used for resizing, scaling, or cropping, then the output resolution width must be a multiple of 8*PPC (Pixel Per Clock). The image_processing kernel can be configured to process 1, 2, or 4 pixels per clock. The higher the PPC value, the higher the throughput. The kernel will always write the data in chunks of 8*PPC. If the output width is not a multiple of 8*PPC, then there might be some garbage data in the right side of the image, depending on the difference between the actual width and the 8*PPC aligned width.