Parameter |
Type |
Expected Values |
Default |
Description |
|---|---|---|---|---|
xclbin-location |
String |
/run/media/mmcblk0p1
/x_plus_ml.xclbin
|
N/A |
Absolute path of the xclbin which contains the image process IP to program the FPGA device based on the device-index property. |
library-name |
String |
libvvas_image_process
_hw_lib.so
|
N/A |
Image process library name ( |
library-config |
String |
{“kernel-name”:
“image_processing
:{image_processing_1}
“}
(for libvvas_image
_process_hw_lib.so)
|
N/A |
Image process library-specific configuration string. The image process library will use this configuration while initializing. |
mean-r |
Float |
0 to 128 |
N/A |
Mean pre-processing value for the Red channel. |
mean-g |
Float |
0 to 128 |
N/A |
Mean pre-processing value for the Green channel. |
mean-b |
Float |
0 to 128 |
N/A |
Mean pre-processing value for the Blue channel. |
scale-r |
Float |
0 to 1 |
N/A |
Scale pre-processing value for the Red channel. |
scale-g |
Float |
0 to 1 |
N/A |
Scale pre-processing value for the Green channel. |
scale-b |
Float |
0 to 1 |
N/A |
Scale pre-processing value for the Blue channel. |
in-mem-bank |
Integer |
0 to 65535 |
0 |
VVAS input memory bank to allocate memory. |
out-mem-bank |
Integer |
0 to 65535 |
0 |
VVAS output memory bank to allocate memory. |
interpolation
-mode
|
Integer |
0 to 2 |
Image process library default interpolation mode |
Used to configure interpolation modes for scaling if supported by the image process library. |
maintain-aspect
-ratio
|
Integer |
0 to 65535 |
0 |
Type of scaling used for the resize operation. Some models require resize with aspect ratio preserved. If not set, default resizing will be applied. |
symmetric-padding |
Integer |
0 to 65535 |
0 |
Applicable if maintain-aspect-ratio is set. If symmetric-padding is enabled, the processed image will be centered with symmetric bars; otherwise, it appears top-left with padding on the bottom or right edges. |
image-pad
-value
|
Integer |
0 to UINT_MAX |
0 |
Pixel value used for padding regions when maintaining aspect ratio. |
quantization |
JSON Object |
See quantization json object |
N/A |
Configuration JSON object for quantization. |
Parameter |
Type |
Expected Values |
Default |
Description |
|---|---|---|---|---|
scale-factor |
Float |
(0.0, 1.0] |
N/A |
Multiplicative factor applied to quantize the pre-processed input before passing to the model. See Quantization and Dequantization. |