MATLAB: Reference to the nonexistent field ‘imdb’.

re-id

I want to change the dataset from 'cuhk03.mat' dataset to 'cuhk03_detected_lomo.mat'(the‘cuhk03_detected_lomo.mat' is a dataset which has been processed with LOMO characteristics ),but it has an error when i run the same program,it shows:
    % Reference to the nonexistent field 'imdb'
    % wrong in train_id_net_res_2stream (line 8)
imdb = imdb.imdb;
I don't understand Why did it go wrong when i just change the dataset. The complete code is as follows：
    % function train_id_net_res_2stream(varargin)
% -------------------------------------------------------------------------





% Part 4.1: prepare the data
% -------------------------------------------------------------------------
% Load character dataset
imdb = load('F:\2016_person_re-ID-master\2016_person_re-ID-master\cuhk03-prepare-eval/cuhk03_detected_lomo.mat') ;
imdb = imdb.imdb;
% -------------------------------------------------------------------------
% Part 4.2: initialize a CNN architecture
% -------------------------------------------------------------------------
net = resnet52_2stream();
net.params(net.getParamIndex('fc751f')).learningRate = 0.01;
net.params(net.getParamIndex('fc751b')).learningRate = 0.2;
net.conserveMemory = true;
net.meta.normalization.averageImage = reshape([105.6920,99.1345,97.9152],1,1,3);
% -------------------------------------------------------------------------
% Part 4.3: train and evaluate the CNN
% -------------------------------------------------------------------------
opts.train.averageImage = net.meta.normalization.averageImage;
opts.train.batchSize = 30;
opts.train.continue = true; 
opts.train.gpus = 1;   %Select gpu card. The gpu id in Matlab start from 1. 
opts.train.prefetch = false ;
opts.train.expDir = './data/resnet52_2stream_drop0.9_new1' ;  % your model will store here
opts.train.learningRate = [0.1*ones(1,70),0.01*ones(1,5)] ;
opts.train.derOutputs = {'objective', 0.5,'objective_2', 0.5,'objective_final', 1} ;
opts.train.weightDecay = 0.0005;
opts.train.numEpochs = numel(opts.train.learningRate) ;
[opts, ~] = vl_argparse(opts.train, varargin) ;
% Call training function in MatConvNet
[net,info] = cnn_train_dag(net, imdb, @getBatch,opts) ;
I'd appreciate it if you could give me some advice.Thank you!
supplement：
When i take the semicolon off the end of the load() line,it shows:
    % imdb = 
    descriptors: [26960x14097 double]
        imgList: {14097x1 cell}
           note: '1_001_1_01.png means the 1st set, the 001th person, the 1st camera, and the…'
Reference to the nonexistent field 'imdb'
    wrong in train_id_net_res_2stream (line 8)
imdb = imdb.imdb;
I'm so sorry that the mat file is so big that i can not attach it with the paper clip icon.

Best Answer

Try
imdb.descriptors
Or
imdb.imgList
Related Question