Dev crop with random size (#1468)

* random size crop proto

* ImagePreprocessImpl::<kCropWithRandomSize>

* clang format

* MaxVal


Former-commit-id: c027432320cc0f03248f9165994150fce058f00a

oneflow/core/record/image.proto

@@ -17,10 +17,25 @@ message ImageCrop {
   required int32 height = 5;
 }
 
+message ImageCropWithRandomSize {
+  message AspectRatioRange {
+    optional float min = 1 [default = 0.75];
+    optional float max = 2 [default = 1.33];
+  }
+  message AreaRange {
+    optional float min = 1 [default = 0.05];
+    optional float max = 2 [default = 1.0];
+  }
+  required AspectRatioRange aspect_ratio_range = 1;
+  required AreaRange area_range = 2;
+  optional int32 max_attempts = 3 [default = 100];
+}
+
 message ImagePreprocess {
   oneof preprocess {
     ImageResize resize = 1;
     ImageCrop crop = 2;
-    ImageMirror mirror = 3;
+    ImageCropWithRandomSize crop_with_random_size = 3;
+    ImageMirror mirror = 4;
   }
 }

oneflow/core/record/image_preprocess.cpp

@@ -2,6 +2,15 @@
 
 namespace oneflow {
 
+namespace {
+
+float GetRandomFloatValue(float min, float max, std::function<int32_t(void)> NextRandomInt) {
+  float ratio = static_cast<float>(NextRandomInt()) / static_cast<float>(MaxVal<int32_t>());
+  return (max - min) * ratio + min;
+}
+
+}  // namespace
+
 void ImagePreprocessImpl<PreprocessCase::kResize>::DoPreprocess(
     cv::Mat* image, const ImagePreprocess& preprocess_conf,
     std::function<int32_t(void)> NextRandomInt) const {
@@ -24,8 +33,10 @@ void ImagePreprocessImpl<PreprocessCase::kCrop>::DoPreprocess(
   CHECK_LE(width, image->cols);
   CHECK_LE(height, image->rows);
   if (crop.random_xy()) {
-    x = NextRandomInt() % (image->cols - width);
-    y = NextRandomInt() % (image->rows - height);
+    int32_t x_max = (image->cols - width);
+    int32_t y_max = (image->rows - height);
+    x = x_max > 0 ? (NextRandomInt() % x_max) : x_max;
+    y = y_max > 0 ? (NextRandomInt() % y_max) : y_max;
   } else {
     CHECK_LE(x, image->cols - width);
     CHECK_LE(y, image->rows - height);
@@ -33,6 +44,38 @@ void ImagePreprocessImpl<PreprocessCase::kCrop>::DoPreprocess(
   *image = (*image)(cv::Rect(x, y, width, height));
 }
 
+void ImagePreprocessImpl<PreprocessCase::kCropWithRandomSize>::DoPreprocess(
+    cv::Mat* image, const ImagePreprocess& preprocess_conf,
+    std::function<int32_t(void)> NextRandomInt) const {
+  CHECK(preprocess_conf.has_crop_with_random_size());
+  const ImageCropWithRandomSize& conf = preprocess_conf.crop_with_random_size();
+  int32_t max_attempts = conf.max_attempts();
+  float area_min = conf.area_range().min();
+  float area_max = conf.area_range().max();
+  float ratio_min = conf.aspect_ratio_range().min();
+  float ratio_max = conf.aspect_ratio_range().max();
+  CHECK_LE(area_min, area_max);
+  CHECK_GT(area_min, 0.0);
+  CHECK_LE(area_max, 1.0);
+  CHECK_LE(ratio_min, ratio_max);
+  CHECK_GT(ratio_min, 0.0);
+  while (max_attempts--) {
+    float area_size = GetRandomFloatValue(area_min, area_max, NextRandomInt) * image->total();
+    float aspect_ratio = GetRandomFloatValue(ratio_min, ratio_max, NextRandomInt);
+    float height_float = sqrt(area_size / aspect_ratio);
+    int32_t height = static_cast<int32_t>(height_float);
+    int32_t width = static_cast<int32_t>(height_float * aspect_ratio);
+    if (width <= image->cols && height <= image->rows) {
+      int32_t x_max = (image->cols - width);
+      int32_t y_max = (image->rows - height);
+      int32_t x = x_max > 0 ? (NextRandomInt() % x_max) : x_max;
+      int32_t y = y_max > 0 ? (NextRandomInt() % y_max) : y_max;
+      *image = (*image)(cv::Rect(x, y, width, height));
+      return;
+    }
+  }
+}
+
 void ImagePreprocessImpl<PreprocessCase::kMirror>::DoPreprocess(
     cv::Mat* image, const ImagePreprocess& preprocess_conf,
     std::function<int32_t(void)> NextRandomInt) const {

oneflow/core/record/image_preprocess.h

@@ -42,6 +42,14 @@ class ImagePreprocessImpl<PreprocessCase::kCrop> final : public ImagePreprocessI
                     std::function<int32_t(void)> NextRandomInt) const override;
 };
 
+template<>
+class ImagePreprocessImpl<PreprocessCase::kCropWithRandomSize> final : public ImagePreprocessIf {
+ public:
+ private:
+  void DoPreprocess(cv::Mat* image, const ImagePreprocess& preprocess_conf,
+                    std::function<int32_t(void)> NextRandomInt) const override;
+};
+
 template<>
 class ImagePreprocessImpl<PreprocessCase::kMirror> final : public ImagePreprocessIf {
  public:
@@ -53,7 +61,8 @@ class ImagePreprocessImpl<PreprocessCase::kMirror> final : public ImagePreproces
 #define PREPROCESS_CASE_SEQ                     \
   OF_PP_MAKE_TUPLE_SEQ(PreprocessCase::kResize) \
   OF_PP_MAKE_TUPLE_SEQ(PreprocessCase::kMirror) \
-  OF_PP_MAKE_TUPLE_SEQ(PreprocessCase::kCrop)
+  OF_PP_MAKE_TUPLE_SEQ(PreprocessCase::kCrop)   \
+  OF_PP_MAKE_TUPLE_SEQ(PreprocessCase::kCropWithRandomSize)
 
 ImagePreprocessIf* GetImagePreprocess(PreprocessCase);