source/use_case/object_detection/src/UseCaseHandler.cc

/*
 * SPDX-FileCopyrightText: Copyright 2022, 2024 Arm Limited and/or its affiliates
 * <open-source-office@arm.com> SPDX-License-Identifier: Apache-2.0
 *
 * Licensed under the Apache License, Version 2.0 (the "License");
 * you may not use this file except in compliance with the License.
 * You may obtain a copy of the License at
 *
 *     http://www.apache.org/licenses/LICENSE-2.0
 *
 * Unless required by applicable law or agreed to in writing, software
 * distributed under the License is distributed on an "AS IS" BASIS,
 * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
 * See the License for the specific language governing permissions and
 * limitations under the License.
 */
#include "UseCaseHandler.hpp"
#include "DetectorPostProcessing.hpp"
#include "DetectorPreProcessing.hpp"
#include "InputFiles.hpp"
#include "UseCaseCommonUtils.hpp"
#include "YoloFastestModel.hpp"
#include "hal.h"
#include "log_macros.h"

#include <cinttypes>

namespace arm {
namespace app {

#if VSI_ENABLED
#include "video_drv.h"                /* Video Driver API. */

    /**
     * @brief Draws a box in the image using the object detection result object.
     *
     * @param[out] imageData    Pointer to the start of the image.
     * @param[in]  width        Image width.
     * @param[in]  height       Image height.
     * @param[in]  result       Object detection result.
     */
    static void DrawBox(uint8_t* imageData,
                        const uint32_t width,
                        const uint32_t height,
                        const OdResults& result)
    {
        UNUSED(height);
        const auto x = result.m_x0;
        const auto y = result.m_y0;
        const auto w = result.m_w;
        const auto h = result.m_h;

        const uint32_t step = width * 3;
        uint8_t* const imStart = imageData + (y * step) + (x * 3);

        uint8_t* dst_0 = imStart;
        uint8_t* dst_1 = imStart + (h * step);

        for (uint32_t i = 0; i < static_cast<uint32_t>(w); ++i) {
            *dst_0 = 255;
            *dst_1 = 255;

            dst_0 += 3;
            dst_1 += 3;
        }

        dst_0 = imStart;
        dst_1 = imStart + (w * 3);

        for (uint32_t j = 0; j < static_cast<uint32_t>(h); ++j) {
            *dst_0 = 255;
            *dst_1 = 255;

            dst_0 += step;
            dst_1 += step;
        }
    }

    void DrawDetectionBoxesVsi(uint8_t* image,
                               const uint32_t imageWidth,
                               const uint32_t imageHeight,
                               const std::vector<OdResults>& results)
    {
        for (const auto& result : results) {
            DrawBox(image, imageWidth, imageHeight, result);
            printf("Detection :: [%d" ", %d"
                   ", %d" ", %d" "]\n",
                   result.m_x0,
                   result.m_y0,
                   result.m_w,
                   result.m_h);
        }
    }

    /* Object detection VSI inference handler. */
    bool ObjectDetectionHandlerVsi(ApplicationContext& ctx)
    {
        /* Image buffer. */
        static uint8_t ImageBuf[IMAGE_DATA_SIZE];
        static uint8_t ImageOut[IMAGE_DATA_SIZE];

        /* Model object creation and initialisation. */
        auto& model = ctx.Get<Model&>("model");

        TfLiteTensor* inputTensor   = model.GetInputTensor(0);
        TfLiteTensor* outputTensor0 = model.GetOutputTensor(0);
        TfLiteTensor* outputTensor1 = model.GetOutputTensor(1);

        if (!inputTensor->dims) {
            printf_err("Invalid input tensor dims\n");
            return false;
        } else if (inputTensor->dims->size < 3) {
            printf_err("Input tensor dimension should be >= 3\n");
            return false;
        }

        TfLiteIntArray* inputShape = model.GetInputShape(0);
        const int inputImgCols = inputShape->data[arm::app::YoloFastestModel::ms_inputColsIdx];
        const int inputImgRows = inputShape->data[arm::app::YoloFastestModel::ms_inputRowsIdx];

        /* Set up pre- and post-processing. */
        arm::app::DetectorPreProcess preProcess =
            arm::app::DetectorPreProcess(inputTensor, true, model.IsDataSigned());

        std::vector<arm::app::OdResults> results;
        const arm::app::object_detection::PostProcessParams postProcessParams{
            inputImgRows,
            inputImgCols,
            arm::app::object_detection::originalImageSize,
            arm::app::object_detection::anchor1,
            arm::app::object_detection::anchor2};
        arm::app::DetectorPostProcess postProcess =
            arm::app::DetectorPostProcess(outputTensor0, outputTensor1, results, postProcessParams);

        const size_t imgSz = inputTensor->bytes < IMAGE_DATA_SIZE ?
                                                                  inputTensor->bytes : IMAGE_DATA_SIZE;

        if (sizeof(ImageBuf) < imgSz) {
            printf_err("Image buffer is insufficient\n");
            return false;
        }

        /* Configure Input Video. */
        if (VideoDrv_Configure(VIDEO_DRV_IN0,
                               arm::app::object_detection::originalImageSize,
                               arm::app::object_detection::originalImageSize,
                               COLOR_RGB888, 24U) != VIDEO_DRV_OK) {
            printf_err("Failed to configure video input\n");
            return false;
        }

        /* Set Input Video buffer. */
        if (VideoDrv_SetBuf(VIDEO_DRV_IN0,  ImageBuf, IMAGE_DATA_SIZE) != VIDEO_DRV_OK) {
            printf_err("Failed to set buffer for video input\n");
            return false;
        }

        /* Set Output Video file (only when using AVH - default: Display) */
        //  if (VideoDrv_SetFile(VIDEO_DRV_OUT0, "output_image.png") != VIDEO_DRV_OK) {
        //      printf_err("Failed to set filename for video output\n");
        //      return 1;
        //  }
        /* Configure Output Video. */
        if (VideoDrv_Configure(VIDEO_DRV_OUT0,
                               arm::app::object_detection::originalImageSize,
                               arm::app::object_detection::originalImageSize,
                               COLOR_RGB888, 24U) != VIDEO_DRV_OK) {
            printf_err("Failed to configure video output\n");
            return false;
        }

        /* Set Output Video buffer. */
        if (VideoDrv_SetBuf(VIDEO_DRV_OUT0, ImageOut, IMAGE_DATA_SIZE) != VIDEO_DRV_OK) {
            printf_err("Failed to set buffer for video output\n");
            return false;
        }

        auto imgCount = ctx.Get<uint32_t>("imgIndex");
        void* imgFrame = nullptr;
        void* outFrame = nullptr;

        while (true) {
#if VSI_IMAGE_INPUT
            if (VideoDrv_SetFile(VIDEO_DRV_IN0, GetFilePath(imgCount)) != VIDEO_DRV_OK) {
                printf_err("Failed to set filename for video input\n");
                return false;
            }
#endif

            VideoDrv_Status_t status;

            results.clear();

            /* Start video capture (single frame). */
            if (VideoDrv_StreamStart(VIDEO_DRV_IN0, VIDEO_DRV_MODE_SINGLE) != VIDEO_DRV_OK) {
                printf_err("Failed to start video capture\n");
                return false;
            }

            /* Wait for video input frame. */
            do {
                status = VideoDrv_GetStatus(VIDEO_DRV_IN0);
            } while (status.buf_empty != 0U);

            /* Get input video frame buffer. */
            imgFrame = VideoDrv_GetFrameBuf(VIDEO_DRV_IN0);

            /* Run the pre-processing, inference and post-processing. */
            if (!preProcess.DoPreProcess(imgFrame, imgSz)) {
                printf_err("Pre-processing failed.\n");
                return false;
            }

            /* Run inference over this image. */
            printf("\rImage %" PRIu32 "; ", ++imgCount);

            if (!model.RunInference()) {
                printf_err("Inference failed.\n");
                return false;
            }

            if (!postProcess.DoPostProcess()) {
                printf_err("Post-processing failed.\n");
                return false;
            }

            /* Release input frame. */
            VideoDrv_ReleaseFrame(VIDEO_DRV_IN0);

            arm::app::DrawDetectionBoxesVsi(static_cast<uint8_t*>(imgFrame), inputImgCols, inputImgRows, results);

            /* Get output video frame buffer. */
            outFrame = VideoDrv_GetFrameBuf(VIDEO_DRV_OUT0);

            /* Copy image frame with detection boxes to output frame buffer. */
            memcpy(outFrame, imgFrame, IMAGE_DATA_SIZE);

            /* Release output frame. */
            VideoDrv_ReleaseFrame(VIDEO_DRV_OUT0);

            /* Start video output (single frame). */
            VideoDrv_StreamStart(VIDEO_DRV_OUT0, VIDEO_DRV_MODE_SINGLE);

            /* Check for end of stream (when using AVH with file as Video input). */
            if (status.eos != 0U) {
                while (VideoDrv_GetStatus(VIDEO_DRV_OUT0).buf_empty == 0U);
                break;
            }
        }

        IncrementAppCtxIfmIdx(ctx, "imgIndex");

        /* De-initialize Video Interface. */
        //VideoDrv_Uninitialize();
        return true;
    }

#endif

    /**
     * @brief           Presents inference results along using the data presentation
     *                  object.
     * @param[in]       results            Vector of detection results to be displayed.
     * @return          true if successful, false otherwise.
     **/
    static bool
    PresentInferenceResult(const std::vector<object_detection::DetectionResult>& results);

    /**
     * @brief           Draw boxes directly on the LCD for all detected objects.
     * @param[in]       results            Vector of detection results to be displayed.
     * @param[in]       imageStartX        X coordinate where the image starts on the LCD.
     * @param[in]       imageStartY        Y coordinate where the image starts on the LCD.
     * @param[in]       imgDownscaleFactor How much image has been downscaled on LCD.
     **/
    static void DrawDetectionBoxes(const std::vector<object_detection::DetectionResult>& results,
                                   uint32_t imgStartX,
                                   uint32_t imgStartY,
                                   uint32_t imgDownscaleFactor);

    /* Object detection inference handler. */
    bool ObjectDetectionHandler(ApplicationContext& ctx, uint32_t imgIndex, bool runAll)
    {
        auto& profiler = ctx.Get<Profiler&>("profiler");

        constexpr uint32_t dataPsnImgDownscaleFactor = 1;
        constexpr uint32_t dataPsnImgStartX          = 10;
        constexpr uint32_t dataPsnImgStartY          = 35;

        constexpr uint32_t dataPsnTxtInfStartX = 20;
        constexpr uint32_t dataPsnTxtInfStartY = 28;

        hal_lcd_clear(COLOR_BLACK);

        auto& model = ctx.Get<Model&>("model");

        /* If the request has a valid size, set the image index. */
        if (imgIndex < NUMBER_OF_FILES) {
            if (!SetAppCtxIfmIdx(ctx, imgIndex, "imgIndex")) {
                return false;
            }
        }
        if (!model.IsInited()) {
            printf_err("Model is not initialised! Terminating processing.\n");
            return false;
        }

        auto initialImgIdx = ctx.Get<uint32_t>("imgIndex");

        TfLiteTensor* inputTensor   = model.GetInputTensor(0);
        TfLiteTensor* outputTensor0 = model.GetOutputTensor(0);
        TfLiteTensor* outputTensor1 = model.GetOutputTensor(1);

        if (!inputTensor->dims) {
            printf_err("Invalid input tensor dims\n");
            return false;
        } else if (inputTensor->dims->size < 3) {
            printf_err("Input tensor dimension should be >= 3\n");
            return false;
        }

        TfLiteIntArray* inputShape = model.GetInputShape(0);

        const int inputImgCols = inputShape->data[YoloFastestModel::ms_inputColsIdx];
        const int inputImgRows = inputShape->data[YoloFastestModel::ms_inputRowsIdx];

        /* Set up pre and post-processing. */
        DetectorPreProcess preProcess = DetectorPreProcess(inputTensor, true, model.IsDataSigned());

        std::vector<object_detection::DetectionResult> results;
        const object_detection::PostProcessParams postProcessParams{
            inputImgRows,
            inputImgCols,
            object_detection::originalImageSize,
            object_detection::anchor1,
            object_detection::anchor2};
        DetectorPostProcess postProcess =
            DetectorPostProcess(outputTensor0, outputTensor1, results, postProcessParams);
        do {
            /* Ensure there are no results leftover from previous inference when running all. */
            results.clear();

            /* Strings for presentation/logging. */
            std::string str_inf{"Running inference... "};

            const uint8_t* currImage = GetImgArray(ctx.Get<uint32_t>("imgIndex"));

            auto dstPtr = static_cast<uint8_t*>(inputTensor->data.uint8);
            const size_t copySz =
                inputTensor->bytes < IMAGE_DATA_SIZE ? inputTensor->bytes : IMAGE_DATA_SIZE;

            /* Run the pre-processing, inference and post-processing. */
            if (!preProcess.DoPreProcess(currImage, copySz)) {
                printf_err("Pre-processing failed.");
                return false;
            }

            /* Display image on the LCD. */
            hal_lcd_display_image(
                (arm::app::object_detection::channelsImageDisplayed == 3) ? currImage : dstPtr,
                inputImgCols,
                inputImgRows,
                arm::app::object_detection::channelsImageDisplayed,
                dataPsnImgStartX,
                dataPsnImgStartY,
                dataPsnImgDownscaleFactor);

            /* Display message on the LCD - inference running. */
            hal_lcd_display_text(
                str_inf.c_str(), str_inf.size(), dataPsnTxtInfStartX, dataPsnTxtInfStartY, false);

            /* Run inference over this image. */
            info("Running inference on image %" PRIu32 " => %s\n",
                 ctx.Get<uint32_t>("imgIndex"),
                 GetFilename(ctx.Get<uint32_t>("imgIndex")));

            if (!RunInference(model, profiler)) {
                printf_err("Inference failed.");
                return false;
            }

            if (!postProcess.DoPostProcess()) {
                printf_err("Post-processing failed.");
                return false;
            }

            /* Erase. */
            str_inf = std::string(str_inf.size(), ' ');
            hal_lcd_display_text(
                str_inf.c_str(), str_inf.size(), dataPsnTxtInfStartX, dataPsnTxtInfStartY, false);

            /* Draw boxes. */
            DrawDetectionBoxes(
                results, dataPsnImgStartX, dataPsnImgStartY, dataPsnImgDownscaleFactor);

#if VERIFY_TEST_OUTPUT
            DumpTensor(modelOutput0);
            DumpTensor(modelOutput1);
#endif /* VERIFY_TEST_OUTPUT */

            if (!PresentInferenceResult(results)) {
                return false;
            }

            profiler.PrintProfilingResult();

            IncrementAppCtxIfmIdx(ctx, "imgIndex");

        } while (runAll && ctx.Get<uint32_t>("imgIndex") != initialImgIdx);

        return true;
    }

    static bool
    PresentInferenceResult(const std::vector<object_detection::DetectionResult>& results)
    {
        hal_lcd_set_text_color(COLOR_GREEN);

        /* If profiling is enabled, and the time is valid. */
        info("Final results:\n");
        info("Total number of inferences: 1\n");

        for (uint32_t i = 0; i < results.size(); ++i) {
            info("%" PRIu32 ") (%f) -> %s {x=%d,y=%d,w=%d,h=%d}\n",
                 i,
                 results[i].m_normalisedVal,
                 "Detection box:",
                 results[i].m_x0,
                 results[i].m_y0,
                 results[i].m_w,
                 results[i].m_h);
        }

        return true;
    }

    static void DrawDetectionBoxes(const std::vector<object_detection::DetectionResult>& results,
                                   uint32_t imgStartX,
                                   uint32_t imgStartY,
                                   uint32_t imgDownscaleFactor)
    {
        uint32_t lineThickness = 1;

        for (const auto& result : results) {
            /* Top line. */
            hal_lcd_display_box(imgStartX + result.m_x0 / imgDownscaleFactor,
                                imgStartY + result.m_y0 / imgDownscaleFactor,
                                result.m_w / imgDownscaleFactor,
                                lineThickness,
                                COLOR_GREEN);
            /* Bot line. */
            hal_lcd_display_box(imgStartX + result.m_x0 / imgDownscaleFactor,
                                imgStartY + (result.m_y0 + result.m_h) / imgDownscaleFactor -
                                    lineThickness,
                                result.m_w / imgDownscaleFactor,
                                lineThickness,
                                COLOR_GREEN);

            /* Left line. */
            hal_lcd_display_box(imgStartX + result.m_x0 / imgDownscaleFactor,
                                imgStartY + result.m_y0 / imgDownscaleFactor,
                                lineThickness,
                                result.m_h / imgDownscaleFactor,
                                COLOR_GREEN);
            /* Right line. */
            hal_lcd_display_box(imgStartX + (result.m_x0 + result.m_w) / imgDownscaleFactor -
                                    lineThickness,
                                imgStartY + result.m_y0 / imgDownscaleFactor,
                                lineThickness,
                                result.m_h / imgDownscaleFactor,
                                COLOR_GREEN);
        }
    }

} /* namespace app */
} /* namespace arm */