I would like to print the labels of traindata / testdata used in classification. Here is the definition of both inputs (using deep4j).
InputSplit[] inputSplit = fileSplit.sample(pathFilter, splitTrainTest, 1 - splitTrainTest);
InputSplit trainData = inputSplit[0];
InputSplit testData = inputSplit[1];
that are then transformed in DataSetIterator like this :
ImageRecordReader recordReader = new ImageRecordReader(height, width, channels, labelMaker);
recordReader.initialize(trainData, null);
trainIter = new RecordReaderDataSetIterator(recordReader, batchSize, 1, numLabels);
Then I want to print how many examples per labels where found in each iterator in this function :
public void print(DataSetIterator iter){
HashMap<String, Integer> hash = new HashMap<String, Integer>();
while(iter.hasNext()){
DataSet example = iter.next();
for(int i = 0 ; i<numLabels ; i++){
if(example.getLabels().getDouble(i)==1.){
String label = example.getLabelName(i);
if(hash.containsKey(label))
hash.put(label, hash.get(label)+1);
else
hash.put(label, 1);
}
}
}
for (String label: hash.keySet()){
System.out.println(" label : " + label.toString() + ", " + hash.get(label) + " examples");
}
}
The issue is that it displays only one example per label, whereas there should much more... And when I don't split my dataset using fileSplit.sample()
the function displays the right number of examples.
Any suggestion ?