train_base_net.m

function train_base_net(varargin)
% load data & label
data = load('./data/train/SDR_youtube_80.mat') ;
label = load('./data/train/HDR_youtube_80.mat') ;
imdb.images.data = data.SDR_data;
imdb.images.label = label.HDR_data;
imdb.images.set = cat(2,ones(1, size(data.SDR_data, 4)-500), 2*ones(1, 500));

% set CNN model
net = net_base();

% set the learning rate and weight decay for biases
% default values are used for filters
for i = 2:2:112
    net.params(i).learningRate = 0.1;
    net.params(i).weightDecay = 0;
end
net.conserveMemory = true;

% options
opts.solver = @adam;
opts.train.batchSize = 16;
opts.train.continue = true; 
opts.train.gpus = 1;
opts.train.prefetch = false ;
opts.train.expDir = './net/net_base' ; 
opts.train.learningRate = 5*1e-7*ones(1, 200);
opts.train.weightDecay = 0.0005;
opts.train.numEpochs = numel(opts.train.learningRate) ;
opts.train.derOutputs = {'objective', 1} ;
[opts, ~] = vl_argparse(opts.train, varargin) ;

% record
if(~isdir(opts.expDir))
    mkdir(opts.expDir);
end

% call training function
[net, info] = cnn_train_dag(net, imdb, @getBatch, opts) ;

function inputs = getBatch(imdb, batch, opts)
image = imdb.images.data(:, :, :, batch) ;
label = imdb.images.label(:, :, :, batch) ;

image = single(image)/255;
label = single(label)/1023;
inputs = {'input', gpuArray(image), 'label', gpuArray(label)};