inst iter

Makefile

@@ -15,7 +15,7 @@ include mshadow/make/mshadow.mk
 # all tge possible warning tread
 WARNFLAGS= -Wall
 CFLAGS = -DMSHADOW_FORCE_STREAM $(WARNFLAGS)
-CFLAGS += -g -O3 -I./mshadow/ -fPIC $(MSHADOW_CFLAGS)
+CFLAGS += -g -O3 -I./mshadow/  -fPIC $(MSHADOW_CFLAGS)
 LDFLAGS = -pthread $(MSHADOW_LDFLAGS)
 NVCCFLAGS = --use_fast_math -g -O3 -ccbin $(CXX) $(MSHADOW_NVCCFLAGS)
 
@@ -33,6 +33,9 @@ else
 	CFLAGS+= -DCXXNET_USE_OPENCV_DECODER=0
 endif
 
+ifeq ($(USE_OPENMP_ITER), 1)
+	CFLAGS += -fopenmp
+endif
 # customize cudnn path
 ifneq ($(USE_CUDNN_PATH), NONE)
 	CFLAGS += -I$(USE_CUDNN_PATH)

make/config.mk

@@ -48,6 +48,8 @@ USE_DIST_PS = 0
 PS_PATH = NONE
 PS_THIRD_PATH = NONE
 
+# use openmp iterator
+USE_OPENMP_ITER = 0
 # the additional link flags you want to add
 ADD_LDFLAGS = -ljpeg
 

src/io/data.cpp

@@ -17,6 +17,7 @@
 #if CXXNET_USE_OPENCV
 #include "iter_thread_imbin-inl.hpp"
 #include "iter_thread_imbin_x-inl.hpp"
+#include "iter_thread_iminst-inl.hpp"
 #include "iter_img-inl.hpp"
 #endif
 
@@ -44,6 +45,11 @@ IIterator<DataBatch> *CreateIterator(const std::vector< std::pair<std::string, s
         it = new BatchAdaptIterator(new AugmentIterator(new ThreadImagePageIteratorX()));
         continue;
       }
+      if (!strcmp(val, "imginst")) {
+        utils::Assert(it == NULL, "image binary can not chain over other iterator");
+        it = new BatchAdaptIterator(new AugmentIterator(new ThreadImageInstIterator(), 1));
+        continue;
+      }
       if (!strcmp(val, "img")) {
         utils::Assert(it == NULL, "image list iterator can not chain over other iterator");
         it = new BatchAdaptIterator(new AugmentIterator(new ImageIterator()));

src/io/image_augmenter-inl.hpp

@@ -92,6 +92,7 @@ class ImageAugmenter {
     // new width and height
     float new_width = std::max(min_img_size_, std::min(max_img_size_, scale * src.cols));
     float new_height = std::max(min_img_size_, std::min(max_img_size_, scale * src.rows));
+    //printf("%f %f %f %f %f %f %f %f %f\n", s, a, b, scale, ratio, hs, ws, new_width, new_height);
     cv::Mat M(2, 3, CV_32F);
     M.at<float>(0, 0) = hs * a - s * b * ws;
     M.at<float>(1, 0) = -b * ws;
@@ -102,7 +103,7 @@ class ImageAugmenter {
     M.at<float>(0, 2) = (new_width - ori_center_width) / 2;
     M.at<float>(1, 2) = (new_height - ori_center_height) / 2;
     cv::warpAffine(src, temp, M, cv::Size(new_width, new_height),
-                     cv::INTER_CUBIC,
+                     cv::INTER_LINEAR,
                      cv::BORDER_CONSTANT,
                      cv::Scalar(fill_value_, fill_value_, fill_value_));
     cv::Mat res = temp;
@@ -150,6 +151,23 @@ class ImageAugmenter {
     return tmpres;
   }
 
+  virtual void Process(unsigned char *dptr, size_t sz, mshadow::TensorContainer<cpu, 3> *p_data,
+                       utils::RandomSampler *prnd) {
+    cv::Mat buf(1, sz, CV_8U, dptr);
+    cv::Mat res = cv::imdecode(buf, 1);
+    res = this->Process(res, prnd);
+    p_data->Resize(mshadow::Shape3(3, res.rows, res.cols));
+    for (index_t i = 0; i < p_data->size(1); ++i) {
+      for (index_t j = 0; j < p_data->size(2); ++j) {
+        cv::Vec3b bgr = res.at<cv::Vec3b>(i, j);
+        (*p_data)[0][i][j] = bgr[2];
+        (*p_data)[1][i][j] = bgr[1];
+        (*p_data)[2][i][j] = bgr[0];
+      }
+    }
+    res.release();
+  }
+
  private:
   // whether skip processing
   inline bool NeedProcess(void) const {

src/io/iter_augment_proc-inl.hpp

@@ -20,8 +20,8 @@ namespace cxxnet {
 /*! \brief create a batch iterator from single instance iterator */
 class AugmentIterator: public IIterator<DataInst> {
 public:
-  AugmentIterator(IIterator<DataInst> *base)
-      : base_(base) {
+  AugmentIterator(IIterator<DataInst> *base, int no_aug=0)
+      : base_(base), no_aug_(no_aug) {
     rand_crop_ = 0;
     rand_mirror_ = 0;
     crop_y_start_ = -1;
@@ -101,7 +101,7 @@ class AugmentIterator: public IIterator<DataInst> {
     out_.index = d.index;
     mshadow::Tensor<cpu, 3> data = d.data;
 #if CXXNET_USE_OPENCV
-    data = aug.Process(data, &rnd);
+    if (!no_aug_) data = aug.Process(data, &rnd);
 #endif
 
     img_.Resize(mshadow::Shape3(data.shape_[0], shape_[1], shape_[2]));
@@ -235,6 +235,7 @@ class AugmentIterator: public IIterator<DataInst> {
   int mirror_;
   /*! \brief whether mean file is ready */
   bool meanfile_ready_;
+  int no_aug_;
   // augmenter
 #if CXXNET_USE_OPENCV
   ImageAugmenter aug;