a) Failed to query Image Process Library caps:
ERROR vvas_xinfer gstvvas_xinfer.cpp:vvas_xinfer_ppe_init:<vvas_xinfer0>
Failed to query Image Process Library caps for library: <library-name>. Check library-name in preprocess-config
ERROR vvas_xinfer gstvvas_xinfer.cpp:gst_vvas_xinfer_change_state:<vvas_xinfer0>
failed to do initialization
Failed to set pipeline to PAUSED.
Possible fix - library-name in preprocess-config refers to a shared library that is not installed or is named incorrectly. Set it to the correct library name:
"preprocess-config": {
"library-name": "libvvas_image_process_hw_lib.so"
}
b) Couldn’t create Image Process instance - invalid or missing library-config:
ERROR vvas_xinfer gstvvas_xinfer.cpp:vvas_xinfer_ppe_init:<vvas_xinfer0>
Couldn't create Image Process instance for library: <library-name>. Possible invalid or missing library-config in preprocess-config
ERROR vvas_xinfer gstvvas_xinfer.cpp:gst_vvas_xinfer_change_state:<vvas_xinfer0>
failed to do initialization
Failed to set pipeline to PAUSED.
Possible fix - library-config in preprocess-config is invalid or missing. Ensure it contains all parameters required by the preprocessing library:
"preprocess-config": {
"library-name": "libvvas_image_process_hw_lib.so",
"library-config": "{\"kernel-name\":\"image_processing:{image_processing_1}\"}",
...
}
c) mean-r is not set / scale-r is not set - pre-process file parse error:
ERROR vvas_xinfer gstvvas_xinfer_helper.cpp:read_ppe_config:<vvas_xinfer0>
mean-r is not set (or: scale-r is not set.)
WARN vvas_xinfer gstvvas_xinfer_helper.cpp:read_ppe_config:<vvas_xinfer0>
error: pre-process file parse error
ERROR: from element .../GstVvas_XInfer:vvas_xinfer0: pre-process file parse error
Failed to set pipeline to PAUSED.
Possible fix - all six normalization fields are mandatory in preprocess-config. Omitting any one causes a parse error at startup. Add all six fields with values matching the model’s training normalization:
"preprocess-config": {
"mean-r": 123.675,
"mean-g": 116.28,
"mean-b": 103.53,
"scale-r": 0.017124,
"scale-g": 0.017507,
"scale-b": 0.017429
}
These values are model-specific. For ResNet50 trained on ImageNet, the preceding JSON block can be used.