多任务深度学习代码

转载自：多任务深度学习
基于Caffe实现多任务学习的小样例

本节在目前广泛使用的深度学习开源框架Caffe的基础上实现多任务深度学习算法所需的多维标签输入。默认的，Caffe中的Data层只支持单维标签，为了支持多维标签，首先修改Caffe中的convert_imageset.cpp以支持多标签：

这样我们就有了多任务的深度学习的基础部分数据输入。为了向上兼容Caffe框架，本文摒弃了部分开源实现增加Data层标签维度选项并修改Data层代码的做法，直接使用两个Data层将数据读入，即分别读入数据和多维标签，接下来介绍对应的网络结构文件prototxt的修改，注意红色的注释部分。

特别的，slice层对多维的标签进行了切分，为每个任务输出了单独的标签。

另外一个值得讨论的是每个任务的权重设置，在本文实践中五个任务设置为等权重loss_weight:0.2。一般的，建议所有任务的权重值相加为1，如果这个数值不设置，可能会导致网络收敛不稳定，这是因为多任务学习中对不同任务的梯度进行累加，导致梯度过大，甚至可能引发参数溢出错误导致网络训练失败。

作者：程程
链接：https://zhuanlan.zhihu.com/p/22190532
来源：知乎
著作权归作者所有。商业转载请联系作者获得授权，非商业转载请注明出处。

多任务损失函数层的网络结构示意图如下图所示：

// This program converts a set of images to a lmdb/leveldb by storing them

// as Datum proto buffers.

// Usage:

// convert_imageset [FLAGS] ROOTFOLDER/ LISTFILE DB_NAME

//

// where ROOTFOLDER is the root folder that holds all the images, and LISTFILE

// should be a list of files as well as their labels, in the format as

// subfolder1/file1.JPEG 7

// ....

//#ifdef MULTILABEL

#include <algorithm>

#include <fstream> // NOLINT(readability/streams)

#include <string>

#include <utility>

#include <vector>

#include "boost/scoped_ptr.hpp"

#include "gflags/gflags.h"

#include "glog/logging.h"

#include "caffe/proto/caffe.pb.h"

#include "caffe/util/db.hpp"

#include "caffe/util/format.hpp"

#include "caffe/util/io.hpp"

#include "caffe/util/rng.hpp"

using namespace caffe; // NOLINT(build/namespaces)

using std::pair;

using boost::scoped_ptr;

DEFINE_bool(gray, false,

"When this option is on, treat images as grayscale ones");

DEFINE_bool(shuffle, false,

"Randomly shuffle the order of images and their labels");

DEFINE_string(backend, "lmdb",

"The backend {lmdb, leveldb} for storing the result");

DEFINE_int32(resize_width, 0, "Width images are resized to");

DEFINE_int32(resize_height, 0, "Height images are resized to");

DEFINE_bool(check_size, false,

"When this option is on, check that all the datum have the same size");

DEFINE_bool(encoded, false,

"When this option is on, the encoded image will be save in datum");

DEFINE_string(encode_type, "",

"Optional: What type should we encode the image as ('png','jpg',...).");

int main(int argc, char** argv) {

#ifdef USE_OPENCV

::google::InitGoogleLogging(argv[0]);

// Print output to stderr (while still logging)

FLAGS_alsologtostderr = 1;

#ifndef GFLAGS_GFLAGS_H_

namespace gflags = google;

#endif

gflags::SetUsageMessage("Convert a set of images to the leveldb/lmdb\n"

"format used as input for Caffe.\n"

"Usage:\n"

" convert_imageset [FLAGS] ROOTFOLDER/ LISTFILE DB_NAME\n"

"The ImageNet dataset for the training demo is at\n"

" http://www.image-net.org/download-images\n");

gflags::ParseCommandLineFlags(&argc, &argv, true);

if (argc < 6) {

gflags::ShowUsageWithFlagsRestrict(argv[0], "tools/convert_imageset");

return 1;

}

const bool is_color = !FLAGS_gray;

const bool check_size = FLAGS_check_size;

const bool encoded = FLAGS_encoded;

const string encode_type = FLAGS_encode_type;

std::ifstream infile(argv[2]);

std::vector<std::pair<std::string, std::vector<float>> > lines;

std::string filename;

std::string label_count_string = argv[5];

int label_count = std::atoi(label_count_string.c_str());

std::vector<float> label(label_count);

while (infile >> filename)

{

for (int i = 0; i < label_count;i++)

{

infile >> label[i];

}

lines.push_back(std::make_pair(filename, label));

}

if (FLAGS_shuffle) {

// randomly shuffle data

LOG(INFO) << "Shuffling data";

shuffle(lines.begin(), lines.end());

}

LOG(INFO) << "A total of " << lines.size() << " images.";

if (encode_type.size() && !encoded)

LOG(INFO) << "encode_type specified, assuming encoded=true.";

int resize_height = std::max<int>(0, FLAGS_resize_height);

int resize_width = std::max<int>(0, FLAGS_resize_width);

// Create new DB

scoped_ptr<db::DB> db_image(db::GetDB(FLAGS_backend));

scoped_ptr<db::DB> db_label(db::GetDB(FLAGS_backend));

db_image->Open(argv[3], db::NEW);

db_label->Open(argv[4], db::NEW);

scoped_ptr<db::Transaction> txn_image(db_image->NewTransaction());

scoped_ptr<db::Transaction> txn_label(db_label->NewTransaction());

// Storing to db

std::string root_folder(argv[1]);

Datum datum_label;

Datum datum_image;

int count = 0;

int data_size_label = 0;

int data_size_image = 0;

bool data_size_initialized = false;

for (int line_id = 0; line_id < lines.size(); ++line_id) {

bool status;

std::string enc = encode_type;

if (encoded && !enc.size()) {

// Guess the encoding type from the file name

string fn = lines[line_id].first;

size_t p = fn.rfind('.');

if (p == fn.npos)

LOG(WARNING) << "Failed to guess the encoding of '" << fn << "'";

enc = fn.substr(p);

std::transform(enc.begin(), enc.end(), enc.begin(), ::tolower);

}

status = ReadImageToDatum(root_folder + lines[line_id].first,

lines[line_id].second[0], resize_height, resize_width, is_color,

enc, &datum_image);

if (status == false) continue;

datum_label.set_height(1);

datum_label.set_width(1);

datum_label.set_channels(label_count);

int count_tmp = datum_label.float_data_size();

for (int index_label = 0; index_label < lines[line_id].second.size(); index_label++)

{

float tmp_float_value = lines[line_id].second[index_label];

datum_label.add_float_data(tmp_float_value);

}

if (check_size) {

if (!data_size_initialized) {

data_size_label = datum_label.channels() * datum_label.height() * datum_label.width();

data_size_image = datum_image.channels() * datum_image.height() * datum_image.width();

data_size_initialized = true;

}

else {

const std::string& data_label = datum_label.data();

CHECK_EQ(data_label.size(), data_size_label) << "Incorrect data field size "

<< data_label.size();

const std::string& data_image = data_image.data();

CHECK_EQ(data_image.size(), data_size_image) << "Incorrect data field size "

<< data_image.size();

}

}

// sequential

string key_str_image = caffe::format_int(line_id, 8) + "_" + lines[line_id].first;

string key_str_label = caffe::format_int(line_id, 8) + "label_" + lines[line_id].first;

// Put in db

string out_label;

string out_image;

CHECK(datum_label.SerializeToString(&out_label));

CHECK(datum_image.SerializeToString(&out_image));

datum_label.clear_float_data();

txn_label->Put(key_str_label, out_label);

txn_image->Put(key_str_image, out_image);

if (++count % 1000 == 0) {

// Commit db

txn_image->Commit();

txn_image.reset(db_image->NewTransaction());

txn_label->Commit();

txn_label.reset(db_label->NewTransaction());

LOG(INFO) << "Processed " << count << " files.";

}

}

// write the last batch

if (count % 1000 != 0) {

txn_label->Commit();

txn_image->Commit();

LOG(INFO) << "Processed " << count << " files.";

}

#else

LOG(FATAL) << "This tool requires OpenCV; compile with USE_OPENCV.";

#endif // USE_OPENCV

return 0;

}

//#endif