nickgillian / grt

gesture recognition toolkit
859 stars 285 forks source link

Failed to find UseExternalRanges #122

Closed alanssoares closed 7 years ago

alanssoares commented 7 years ago

Hi, I'm trying test my dataset in your toolkit using some algorithms. When I run my training using a file created (my dataset) according with your description .grt file, it shows a message:

[ERROR TSCD] loadDatasetFromFile(std::string filename) - Failed to find UseExternalRanges! Failed to load training data!

My dataset:

GRT_LABELLED_TIME_SERIES_CLASSIFICATION_DATA_FILE_V1.0 DatasetName: UFBAGRDataset InfoText: This dataset contains 7 gestures (g1,g2,g3,g4,g5,g6,g7), totalizing 1099 executions. NumDimensions: 3 TotalNumTrainingExamples: 600 NumberOfClasses: 7 ClassIDsAndCounters: g1 182 g2 63 g3 220 g4 135 g5 194 g6 147 g7 158 UseExternalRanges: 0 LabelledTimeSeriesTrainingData: ****TIME_SERIES**** ClassID: g1 TimeSeriesLength: 33 TimeSeriesData: 0.319118 0 1 0.309124 0.00507767 0.993363 0.31764 0.0116771 0.980371 0.313818 0.0317053 0.979665 0.303218 0.0401768 0.972927 ...

I also have some doubts:

alanssoares commented 7 years ago

Well, I found the problem. The documentation doesn't have information about the usage of the different versions of the dataset description file like:

GRT_LABELLED_TIME_SERIES_CLASSIFICATION_DATA_FILE_V1.0 GRT_LABELLED_CLASSIFICATION_DATA_FILE_V1.0

Answering the question that myself asked:

Q1 - Can I use 3D vector of float values in time series?

R: Yes, you need set the NumDimensions: 3

Q2 - Can I use the same description file with others algorithms/models (HMM,SVM)?

R: No, you can't. The description file depends of the algorithm.

Q3 - My gestures are performed with two hands, so I created the time series first putting all left hand positions and after the right. That's right?

R: Well, I don't know. I hope that someone answer this question.

Anyway, I hope that this can help.