Performance

Benchmark mode

The application you built earlier includes a benchmark mode that runs the core processing loop multiple times in a hot loop:

• ai-camera-pipelines/bin/cinematic_mode_benchmark
• ai-camera-pipelines/bin/low_light_image_enhancement_benchmark

These benchmarks demonstrate the performance improvements enabled by KleidiCV and KleidiAI:

• KleidiCV enhances OpenCV performance with computation kernels optimized for Arm processors.

• KleidiAI accelerates TFLite + XNNPack inference using AI-optimized micro-kernels tailored for Arm CPUs.

Performances with KleidiCV and KleidiAI

By default, the OpenCV library is built with KleidiCV support, and TFLite+xnnpack is built with KleidiAI support.

You can run the benchmarks using the applications you built earlier.

Run the first benchmark:

    

        
        
bin/cinematic_mode_benchmark 20 resources/depth_and_saliency_v3_2_assortedv2_w_augment_mobilenetv2_int8_only_ptq.tflite

    

The output is similar to:

    

        
        INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Total run time over 20 iterations: 2023.39 ms

        
    

Run the second benchmark:

    

        
        
bin/low_light_image_enhancement_benchmark 20 resources/HDRNetLIME_lr_coeffs_v1_1_0_mixed_low_light_perceptual_l2_loss_int8_only_ptq.tflite

    

The output is similar to:

    

        
        INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Total run time over 20 iterations: 54.3546 ms

        
    

From these results, you can see that:

• cinematic_mode_benchmark performed 20 iterations in 1985.99 ms.
• low_light_image_enhancement_benchmark performed 20 iterations in 52.3448 ms.

Benchmark results without KleidiCV and KleidiAI

To measure the performance without these optimizations, recompile the pipelines using the following flags in your CMake command:

    

        
        
-DENABLE_KLEIDICV:BOOL=OFF -DXNNPACK_ENABLE_KLEIDIAI:BOOL=OFF

    

Re-run the first benchmark:

    

        
        
bin/cinematic_mode_benchmark 20 resources/depth_and_saliency_v3_2_assortedv2_w_augment_mobilenetv2_int8_only_ptq.tflite

    

The new output is similar to:

    

        
        INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Total run time over 20 iterations: 2029.25 ms

        
    

Re-run the second benchmark:

    

        
        
bin/low_light_image_enhancement_benchmark 20 resources/HDRNetLIME_lr_coeffs_v1_1_0_mixed_low_light_perceptual_l2_loss_int8_only_ptq.tflite

    

The new output is similar to:

    

        
        INFO: Created TensorFlow Lite XNNPACK delegate for CPU.
Total run time over 20 iterations: 79.431 ms

        
    

Comparison table

As shown, the background blur pipeline (cinematic_mode_benchmark) gains only a small improvement, while the low-light enhancement pipeline sees a significant ~30% performance uplift when KleidiCV and KleidiAI are enabled.

Future performance uplift with SME2

A major benefit of using KleidiCV and KleidiAI is that they can automatically leverage new Arm architecture features - such as SME2 (Scalable Matrix Extension v2) - without requiring changes to your application code.

As KleidiCV and KleidiAI operate as performance abstraction layers, any future hardware instruction support can be utilized by simply rebuilding the application. This enables better performance on newer processors without additional engineering effort.