Minor changes: delete TODO.txt; ran/fixed lint; better formatting, etc.

include/caffe/blob.hpp

@@ -138,13 +138,14 @@ class Blob {
   inline int channels() const { return LegacyShape(1); }
   /// @brief Deprecated legacy shape accessor length: use shape(2) instead.
   inline int length() const { return (num_axes() == 5) ? LegacyShape(2) : 1; }
-  //inline int length() const { return LegacyShape(2); }
   /// @brief Deprecated legacy shape accessor height: use shape(3) instead.
-  inline int height() const { return (num_axes() == 5) ? LegacyShape(3) : LegacyShape(2); }
-  //inline int height() const { return LegacyShape(3); }
+  inline int height() const {
+    return (num_axes() == 5) ? LegacyShape(3) : LegacyShape(2);
+  }
   /// @brief Deprecated legacy shape accessor width: use shape(4) instead.
-  inline int width() const { return (num_axes() == 5) ? LegacyShape(4) : LegacyShape(3); }
-  //inline int width() const { return LegacyShape(4); }
+  inline int width() const {
+    return (num_axes() == 5) ? LegacyShape(4) : LegacyShape(3);
+  }
   inline int LegacyShape(int index) const {
     CHECK_LE(num_axes(), 5)
         << "Cannot use legacy accessors on Blobs with > 5 axes.";

include/caffe/layers/video_data_layer.hpp

@@ -15,8 +15,7 @@
 // an extension the std::pair which used to store image filename and
 // its label (int). now, a frame number associated with the video filename
 // is needed (second param) to fully represent a video segment
-struct triplet
-{
+struct triplet {
   std::string first;
   int second, third;
 };

include/caffe/util/im2col.hpp

@@ -7,7 +7,7 @@ template <typename Dtype>
 void im2col_nd_cpu(const Dtype* data_im, const int num_spatial_axes,
     const int* im_shape, const int* col_shape,
     const int* kernel_shape, const int* pad, const int* stride,
-    const int* dilation, Dtype* data_col, const bool forced_3d=false);
+    const int* dilation, Dtype* data_col, const bool forced_3d = false);
 
 template <typename Dtype>
 void im2col_cpu(const Dtype* data_im, const int channels,
@@ -20,7 +20,7 @@ template <typename Dtype>
 void col2im_nd_cpu(const Dtype* data_col, const int num_spatial_axes,
     const int* im_shape, const int* col_shape,
     const int* kernel_shape, const int* pad, const int* stride,
-    const int* dilation, Dtype* data_im, const bool forced_3d=false);
+    const int* dilation, Dtype* data_im, const bool forced_3d = false);
 
 template <typename Dtype>
 void col2im_cpu(const Dtype* data_col, const int channels,

include/caffe/util/io.hpp

@@ -5,6 +5,7 @@
 #include <iomanip>
 #include <iostream>  // NOLINT(readability/streams)
 #include <string>
+#include <vector>
 
 #include "google/protobuf/message.h"
 

scripts/cpp_lint.py

@@ -1595,10 +1595,10 @@ def CheckCaffeAlternatives(filename, clean_lines, linenum, error):
 def CheckCaffeDataLayerSetUp(filename, clean_lines, linenum, error):
   """Except the base classes, Caffe DataLayer should define DataLayerSetUp
      instead of LayerSetUp.
-     
+
   The base DataLayers define common SetUp steps, the subclasses should
   not override them.
-  
+
   Args:
     filename: The name of the current file.
     clean_lines: A CleansedLines instance containing the file.
@@ -1610,6 +1610,7 @@ def CheckCaffeDataLayerSetUp(filename, clean_lines, linenum, error):
   if ix >= 0 and (
        line.find('void DataLayer<Dtype>::LayerSetUp') != -1 or
        line.find('void ImageDataLayer<Dtype>::LayerSetUp') != -1 or
+       line.find('void VideoDataLayer<Dtype>::LayerSetUp') != -1 or
        line.find('void MemoryDataLayer<Dtype>::LayerSetUp') != -1 or
        line.find('void WindowDataLayer<Dtype>::LayerSetUp') != -1):
       error(filename, linenum, 'caffe/data_layer_setup', 2,
@@ -1622,6 +1623,7 @@ def CheckCaffeDataLayerSetUp(filename, clean_lines, linenum, error):
        line.find('void Base') == -1 and
        line.find('void DataLayer<Dtype>::DataLayerSetUp') == -1 and
        line.find('void ImageDataLayer<Dtype>::DataLayerSetUp') == -1 and
+       line.find('void VideoDataLayer<Dtype>::DataLayerSetUp') == -1 and
        line.find('void MemoryDataLayer<Dtype>::DataLayerSetUp') == -1 and
        line.find('void WindowDataLayer<Dtype>::DataLayerSetUp') == -1):
       error(filename, linenum, 'caffe/data_layer_setup', 2,

src/caffe/blob.cpp

@@ -34,15 +34,8 @@ void Blob<Dtype>::Reshape(const int num, const int channels, const int height,
 
 template <typename Dtype>
 void Blob<Dtype>::Reshape(const vector<int>& shape) {
-  //std::cout << "-----------------------------"<<std::endl;
-  //for (int i=0; i<shape.size(); ++i)
-  //  std::cout<<"shape["<<i<<"]="<<shape[i]<<std::endl;
   vector<int> new_shape = shape;
   const bool force_3d_ = (shape.size() == 4) && false;
-  //if (force_3d_)
-  //  std::cout<<"!!!!! force_3d_="<<force_3d_<<" !!!!!"<<std::endl;
-  //else
-  //  std::cout<<"force_3d_="<<force_3d_<<std::endl;
   if (force_3d_)
     new_shape.insert(new_shape.begin() + 2, 1);
   CHECK_LE(new_shape.size(), kMaxBlobAxes);
@@ -64,9 +57,6 @@ void Blob<Dtype>::Reshape(const vector<int>& shape) {
     data_.reset(new SyncedMemory(capacity_ * sizeof(Dtype)));
     diff_.reset(new SyncedMemory(capacity_ * sizeof(Dtype)));
   }
-  //std::cout << "----------"<<std::endl;
-  //for (int i=0; i<shape_.size(); ++i)
-  //  std::cout<<"shape_["<<i<<"]="<<shape_[i]<<std::endl;
 }
 
 template <typename Dtype>
@@ -441,8 +431,7 @@ bool Blob<Dtype>::ShapeEquals(const BlobProto& other) {
            LegacyShape(-3) == other.length() &&
            LegacyShape(-2) == other.height() &&
            LegacyShape(-1) == other.width();
-  }
-  else if ((other.has_num() || other.has_channels() || other.has_height() ||
+  } else if ((other.has_num() || other.has_channels() || other.has_height() ||
             other.has_width()) && !other.has_length()) {
     // Using deprecated 4D Blob dimensions --
     // shape is (num, channels, height, width).
@@ -498,10 +487,11 @@ void Blob<Dtype>::CopyFrom(const Blob& source, bool copy_diff, bool reshape) {
 
 template <typename Dtype>
 void Blob<Dtype>::FromProto(const BlobProto& proto, bool reshape) {
+  bool is_legacy_C3D_proto = false;
   if (reshape) {
     vector<int> shape;
-    if ((proto.has_num() || proto.has_channels() ||
-        proto.has_height() || proto.has_width()) && proto.has_length()) {
+    if (proto.has_num() && proto.has_channels() && proto.has_height() &&
+        proto.has_width() && proto.has_length()) {
       // Using deprecated 5D Blob dimensions --
       // shape is (num, channels, length, height, width).
       shape.resize(5);
@@ -510,18 +500,31 @@ void Blob<Dtype>::FromProto(const BlobProto& proto, bool reshape) {
       shape[2] = proto.length();
       shape[3] = proto.height();
       shape[4] = proto.width();
-    }
     // backward compatibility
-    else if ((proto.has_num() || proto.has_channels() ||
-        proto.has_height() || proto.has_width()) && !proto.has_length()) {
-      // Using deprecated 4D Blob dimensions --
-      // shape is (num, channels, height, width).
-      shape.resize(5);
-      shape[0] = proto.num();
-      shape[1] = proto.channels();
-      shape[2] = 1;
-      shape[3] = proto.height();
-      shape[4] = proto.width();
+    } else if (proto.has_num() && proto.has_channels() && proto.has_height() &&
+               proto.has_width() && !proto.has_length()) {
+      // a hacky way to support old proto generated by facebook/C3D
+      const int forgotten_dim_size = proto.diff_size() / (proto.num() *
+        proto.channels() * proto.height() * proto.width());
+      is_legacy_C3D_proto = (forgotten_dim_size > 1);
+      LOG(INFO) << "legacy C3D proto detected";
+      if (is_legacy_C3D_proto) {
+        shape.resize(5);
+        shape[0] = proto.num();
+        shape[1] = proto.channels();
+        shape[2] = proto.height();
+        shape[3] = proto.width();
+        shape[4] = forgotten_dim_size;
+      } else {
+        // Using deprecated 4D Blob dimensions --
+        // shape is (num, channels, height, width).
+        shape.resize(5);
+        shape[0] = proto.num();
+        shape[1] = proto.channels();
+        shape[2] = 1;
+        shape[3] = proto.height();
+        shape[4] = proto.width();
+      }
     } else {
       shape.resize(proto.shape().dim_size());
       for (int i = 0; i < proto.shape().dim_size(); ++i) {
@@ -540,9 +543,17 @@ void Blob<Dtype>::FromProto(const BlobProto& proto, bool reshape) {
       data_vec[i] = proto.double_data(i);
     }
   } else {
-    CHECK_EQ(count_, proto.data_size());
-    for (int i = 0; i < count_; ++i) {
-      data_vec[i] = proto.data(i);
+    // normal case
+    if (!is_legacy_C3D_proto) {
+      CHECK_EQ(count_, proto.data_size());
+      for (int i = 0; i < count_; ++i) {
+        data_vec[i] = proto.data(i);
+      }
+    // binary proto file created by legacy C3D code
+    } else {
+      for (int i = 0; i < count_; ++i) {
+        data_vec[i] = proto.diff(i);
+      }
     }
   }
   if (proto.double_diff_size() > 0) {
@@ -551,7 +562,7 @@ void Blob<Dtype>::FromProto(const BlobProto& proto, bool reshape) {
     for (int i = 0; i < count_; ++i) {
       diff_vec[i] = proto.double_diff(i);
     }
-  } else if (proto.diff_size() > 0) {
+  } else if (proto.diff_size() > 0 && !is_legacy_C3D_proto) {
     CHECK_EQ(count_, proto.diff_size());
     Dtype* diff_vec = mutable_cpu_diff();
     for (int i = 0; i < count_; ++i) {

src/caffe/layers/base_conv_layer.cpp

@@ -17,13 +17,11 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
   channel_axis_ = bottom[0]->CanonicalAxisIndex(conv_param.axis());
   const int num_axes = bottom[0]->num_axes();
   int first_spatial_axis;
-  if (num_axes == 5 && channel_axis_ == 1 &&
-      bottom[0]->shape(2) == 1) {
+  if (num_axes == 5 && channel_axis_ == 1 && bottom[0]->shape(2) == 1) {
     forced_3d_ = true;
-    first_spatial_axis = 3; // not 2
-    num_spatial_axes_ = 2;  // not 3
-  }
-  else {
+    first_spatial_axis = 3;   // not 2
+    num_spatial_axes_ = 2;    // not 3
+  } else {
     forced_3d_ = false;
     first_spatial_axis = channel_axis_ + 1;
     num_spatial_axes_ = num_axes - first_spatial_axis;
@@ -47,22 +45,21 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
       CHECK(num_kernel_dims == 1 ||
             num_kernel_dims == num_spatial_axes_ ||
             num_kernel_dims == num_spatial_axes_ + 1)
-          << "kernel_size must be specified once, or once per spatial dimension "
-          << "(kernel_size specified " << num_kernel_dims << " times; "
-          << num_spatial_axes_ << " spatial dims).";
+          << "kernel_size must be specified once, or once per spatial "
+          << "dimension (kernel_size specified " << num_kernel_dims
+          << " times; " << num_spatial_axes_ << " spatial dims).";
     } else {
       CHECK(num_kernel_dims == 1 || num_kernel_dims == num_spatial_axes_)
-          << "kernel_size must be specified once, or once per spatial dimension "
-          << "(kernel_size specified " << num_kernel_dims << " times; "
-          << num_spatial_axes_ << " spatial dims).";
+          << "kernel_size must be specified once, or once per spatial "
+          << "dimension (kernel_size specified " << num_kernel_dims
+          << " times; " << num_spatial_axes_ << " spatial dims).";
     }
     if (num_kernel_dims == 1) {
       for (int i = 0; i < num_spatial_axes_; ++i) {
         kernel_shape_data[i] =
             conv_param.kernel_size(0);
       }
-    }
-    else if (num_kernel_dims == num_spatial_axes_) {
+    } else if (num_kernel_dims == num_spatial_axes_) {
       for (int i = 0; i < num_spatial_axes_; ++i) {
         kernel_shape_data[i] =
             conv_param.kernel_size(i);
@@ -71,7 +68,7 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
     if (num_kernel_dims == num_spatial_axes_ + 1) {
       for (int i = 0; i < num_spatial_axes_; ++i) {
         kernel_shape_data[i] =
-            conv_param.kernel_size(i + 1); // ignore the first kernel_size
+            conv_param.kernel_size(i + 1);   // ignore the first kernel_size
       }
     }
   }
@@ -113,7 +110,7 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
       else if (num_stride_dims == num_spatial_axes_ )
         stride_data[i] = conv_param.stride(i);
       else if (num_stride_dims == num_spatial_axes_ + 1)
-        stride_data[i] = conv_param.stride(i + 1); // ignore the first one
+        stride_data[i] = conv_param.stride(i + 1);   // ignore the first one
       CHECK_GT(stride_data[i], 0) << "Stride dimensions must be nonzero.";
     }
   }
@@ -152,7 +149,7 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
       else if (num_pad_dims == num_spatial_axes_ )
         pad_data[i] = conv_param.pad(i);
       else if (num_pad_dims == num_spatial_axes_ + 1)
-        pad_data[i] = conv_param.pad(i + 1); // ignore the first one
+        pad_data[i] = conv_param.pad(i + 1);   // ignore the first one
     }
   }
   // Setup dilation dimensions (dilation_).
@@ -182,7 +179,7 @@ void BaseConvolutionLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
     else if (num_dilation_dims == num_spatial_axes_ )
       dilation_data[i] = conv_param.dilation(i);
     else if (num_dilation_dims == num_spatial_axes_ + 1)
-      dilation_data[i] = conv_param.dilation(i + 1); // ignore the first one
+      dilation_data[i] = conv_param.dilation(i + 1);   // ignore the first one
   }
   // Special case: im2col is the identity for 1x1 convolution with stride 1
   // and no padding, so flag for skipping the buffer and transformation.
@@ -305,7 +302,8 @@ void BaseConvolutionLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
     if (reverse_dimensions()) {
       conv_input_shape_data[i] = top[0]->shape(channel_axis_ + i + forced_3d_);
     } else {
-      conv_input_shape_data[i] = bottom[0]->shape(channel_axis_ + i + forced_3d_);
+      conv_input_shape_data[i] = bottom[0]->shape(channel_axis_ + i +
+          forced_3d_);
     }
   }
   // The im2col result buffer will only hold one image at a time to avoid

src/caffe/layers/cudnn_conv_layer.cpp

@@ -98,8 +98,6 @@ void CuDNNConvolutionLayer<Dtype>::Reshape(
   bottom_offset_ = this->bottom_dim_ / this->group_;
   top_offset_ = this->top_dim_ / this->group_;
   const bool forced_3d = this->forced_3d_;
-  //std::cout << "cudnn_conv_layer.cpp: num_spatial_axes="<<this->num_spatial_axes_<<std::endl;
-  //std::cout << "cudnn_conv_layer.cpp: forced_3d="<<forced_3d<<std::endl;
   const int height = bottom[0]->shape(this->channel_axis_ + 1 + forced_3d);
   const int width = bottom[0]->shape(this->channel_axis_ + 2 + forced_3d);
   const int height_out = top[0]->shape(this->channel_axis_ + 1 + forced_3d);
@@ -110,8 +108,6 @@ void CuDNNConvolutionLayer<Dtype>::Reshape(
   const int* stride_data = this->stride_.cpu_data();
   const int stride_h = stride_data[0];
   const int stride_w = stride_data[1];
-  //std::cout << "cudnn_conv_layer.cpp: h="<<height<<", w="<<width<<", h_o="<<height_out<<
-  //             ", w_o="<<width_out<<", pad_h="<<pad_h <<std::endl;
 
   // Specify workspace limit for kernels directly until we have a
   // planning strategy and a rewrite of Caffe's GPU memory mangagement

src/caffe/layers/im2col_layer.cpp

@@ -18,10 +18,8 @@ void Im2colLayer<Dtype>::LayerSetUp(const vector<Blob<Dtype>*>& bottom,
     forced_3d_ = true;
   else
     forced_3d_ = false;
-  //std::cout << "forced_3d_=" << forced_3d_ << std::endl;
   const int first_spatial_dim = channel_axis_ + 1 + forced_3d_;
   num_spatial_axes_ = input_num_dims - first_spatial_dim;
-  //std::cout << "num_spatial_axes_=" << num_spatial_axes_ << std::endl;
   CHECK_GE(num_spatial_axes_, 1);
   vector<int> dim_blob_shape(1, num_spatial_axes_);
   // Setup filter kernel dimensions (kernel_shape_).
@@ -128,11 +126,6 @@ void Im2colLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
         / stride_data[i] + 1;
     top_shape[channel_axis_ + i + 1 + forced_3d_] = output_dim;
   }
-  //{
-  //std::cout << "----------------------" << std::endl;
-  //for (int i=0; i<top_shape.size(); ++i)
-  //  std::cout << "top_shape["<<i<<"]="<<top_shape[i]<<std::endl;
-  //}
   top[0]->Reshape(top_shape);
   num_ = bottom[0]->count(0, channel_axis_);
   bottom_dim_ = bottom[0]->count(channel_axis_);

src/caffe/layers/pooling_layer.cpp

@@ -86,7 +86,6 @@ void PoolingLayer<Dtype>::Reshape(const vector<Blob<Dtype>*>& bottom,
   channels_ = bottom[0]->channels();
   height_ = bottom[0]->height();
   width_ = bottom[0]->width();
-  //std::cout<<"c="<<channels_<<",h="<<height_<<",w="<<width_<<std::endl;
   if (global_pooling_) {
     kernel_h_ = bottom[0]->height();
     kernel_w_ = bottom[0]->width();