Suggestions on Optimizing DVPP+AIPP+Model Inference Accuracy (Atlas 200/300/500 Inference Product )
In the event of DVPP+AIPP+model inference cascading, improper arguments passed to API calls or improper AIPP parameters in model conversion will lead to abnormal input and output connections between the cascaded pipelines and finally unsatisfactory inference results.
For details about accuracy optimization suggestions, typical cases, and sample code, see JPEGD+VPC+Model Inference (Atlas 200/300/500 Inference Product).
Parent topic: Accuracy Improvement Suggestions for Model Inference