MLECO-2458 and MLECO-2476 [Fix] VWW IFM quant step

* Changed image->cc conversion to be similar with preprocessing
  of img_class and vww models: images are scaled maintaing the
  aspect ration and then the centre crop of the correct size
  is taken.
* VWW applies input quantization info to the int8 image
  (prior converted to [0,1] float range).
* Changed adult_blur to a image without person.
* Fix menu print when selecting a specific ifm to run
  (Select message was displayed after typing something)

Change-Id: Ie6cde7ab4835ea842667b87397458a5d32131df3

8 files changed, 20 insertions, 6 deletions

diff --git a/source/use_case/ad/src/MainLoop.cc b/source/use_case/ad/src/MainLoop.cc
index 3c2f9cc..a323610 100644
--- a/source/use_case/ad/src/MainLoop.cc
+++ b/source/use_case/ad/src/MainLoop.cc
@@ -90,6 +90,7 @@ void main_loop(hal_platform& platform)
             case MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the data index [0, %d]: ",
                        NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto audioIndex = static_cast<uint32_t>(
                         arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyVibrationHandler(caseContext,
diff --git a/source/use_case/asr/src/MainLoop.cc b/source/use_case/asr/src/MainLoop.cc
index 42bb08e..87145f4 100644
--- a/source/use_case/asr/src/MainLoop.cc
+++ b/source/use_case/asr/src/MainLoop.cc
@@ -133,6 +133,7 @@ void main_loop(hal_platform& platform)
             case MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the audio clip index [0, %d]: ",
                        NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto clipIndex = static_cast<uint32_t>(
                                     arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyAudioHandler(caseContext,
diff --git a/source/use_case/img_class/src/MainLoop.cc b/source/use_case/img_class/src/MainLoop.cc
index 79f6018..ea9f14a 100644
--- a/source/use_case/img_class/src/MainLoop.cc
+++ b/source/use_case/img_class/src/MainLoop.cc
@@ -68,6 +68,7 @@ void main_loop(hal_platform& platform)
                 break;
             case common::MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the image index [0, %d]: ", NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto imgIndex = static_cast<uint32_t>(arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyImageHandler(caseContext, imgIndex, false);
                 break;
diff --git a/source/use_case/img_class/src/UseCaseHandler.cc b/source/use_case/img_class/src/UseCaseHandler.cc
index 66df1da..effc06f 100644
--- a/source/use_case/img_class/src/UseCaseHandler.cc
+++ b/source/use_case/img_class/src/UseCaseHandler.cc
@@ -97,7 +97,7 @@ namespace app {
 
             /* Display this image on the LCD. */
             platform.data_psn->present_data_image(
-                (uint8_t*) inputTensor->data.data,
+                static_cast<uint8_t *>(inputTensor->data.data),
                 nCols, nRows, nChannels,
                 dataPsnImgStartX, dataPsnImgStartY, dataPsnImgDownscaleFactor);
 
diff --git a/source/use_case/kws/src/MainLoop.cc b/source/use_case/kws/src/MainLoop.cc
index 80254d0..c683e71 100644
--- a/source/use_case/kws/src/MainLoop.cc
+++ b/source/use_case/kws/src/MainLoop.cc
@@ -95,6 +95,7 @@ void main_loop(hal_platform& platform)
                 break;
             case MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the audio clip index [0, %d]: ", NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto clipIndex = static_cast<uint32_t>(arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyAudioHandler(caseContext, clipIndex, false);
                 break;
diff --git a/source/use_case/kws_asr/src/MainLoop.cc b/source/use_case/kws_asr/src/MainLoop.cc
index ed9034f..d5a2c2b 100644
--- a/source/use_case/kws_asr/src/MainLoop.cc
+++ b/source/use_case/kws_asr/src/MainLoop.cc
@@ -161,6 +161,7 @@ void main_loop(hal_platform& platform)
             case MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the audio clip index [0, %d]: ",
                        NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto clipIndex = static_cast<uint32_t>(
                         arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyAudioHandler(caseContext,
diff --git a/source/use_case/vww/src/MainLoop.cc b/source/use_case/vww/src/MainLoop.cc
index f026cc2..b29238c 100644
--- a/source/use_case/vww/src/MainLoop.cc
+++ b/source/use_case/vww/src/MainLoop.cc
@@ -67,6 +67,7 @@ void main_loop(hal_platform &platform)
                 break;
             case common::MENU_OPT_RUN_INF_CHOSEN: {
                 printf("    Enter the image index [0, %d]: ", NUMBER_OF_FILES-1);
+                fflush(stdout);
                 auto imgIndex = static_cast<uint32_t>(arm::app::ReadUserInputAsInt(platform));
                 executionSuccessful = ClassifyImageHandler(caseContext, imgIndex, false);
                 break;
diff --git a/source/use_case/vww/src/UseCaseHandler.cc b/source/use_case/vww/src/UseCaseHandler.cc
index fb2e837..d384032 100644
--- a/source/use_case/vww/src/UseCaseHandler.cc
+++ b/source/use_case/vww/src/UseCaseHandler.cc
@@ -21,6 +21,8 @@
 #include "UseCaseCommonUtils.hpp"
 #include "hal.h"
 
+#include <algorithm>
+
 namespace arm {
 namespace app {
 
@@ -94,13 +96,19 @@ namespace app {
 
             /* Display this image on the LCD. */
             platform.data_psn->present_data_image(
-                (uint8_t *) inputTensor->data.data,
+                static_cast<uint8_t *>(inputTensor->data.data),
                 nCols, nRows, nChannels,
                 dataPsnImgStartX, dataPsnImgStartY, dataPsnImgDownscaleFactor);
 
-            /* If the data is signed. */
-            if (model.IsDataSigned()) {
-                image::ConvertImgToInt8(inputTensor->data.data, inputTensor->bytes);
+            /* Vww model preprocessing is image conversion from uint8 to [0,1] float values,
+             * then quantize them with input quantization info. */
+            QuantParams inQuantParams = GetTensorQuantParams(inputTensor);
+
+            auto* req_data = static_cast<uint8_t *>(inputTensor->data.data);
+            auto* signed_req_data = static_cast<int8_t *>(inputTensor->data.data);
+            for (size_t i = 0; i < inputTensor->bytes; i++) {
+                auto i_data_int8 = static_cast<int8_t>(((static_cast<float>(req_data[i]) / 255.0f) / inQuantParams.scale) + inQuantParams.offset);
+                signed_req_data[i] = std::min<int8_t>(INT8_MAX, std::max<int8_t>(i_data_int8, INT8_MIN));
             }
 
             /* Display message on the LCD - inference running. */
@@ -159,7 +167,7 @@ namespace app {
         const uint32_t nChannels = (inputTensor->dims->size == 4) ? inputTensor->dims->data[3] : 1;
 
         const uint8_t* srcPtr = get_img_array(imIdx);
-        auto* dstPtr = (uint8_t*)inputTensor->data.data;
+        auto* dstPtr = static_cast<uint8_t *>(inputTensor->data.data);
         if (1 == nChannels) {
             /**
              * Visual Wake Word model accepts only one channel =>