Skip to content

Reading vectors

I'm using the application VectorClassifier through the command line launcher to predict samples.

Here is the otb report :

2019-10-31 16:23:32 (INFO) VectorClassifier: Default RAM limit for OTB is 256 MB
2019-10-31 16:23:32 (INFO) VectorClassifier: GDAL maximum cache size is 128 MB
2019-10-31 16:23:32 (INFO) VectorClassifier: OTB will use at most 12 threads
2019-10-31 16:23:36 (INFO) VectorClassifier: Model loaded
2019-10-31 16:54:33 (INFO) VectorClassifier: mean used: [0, ..., 0]
2019-10-31 16:54:33 (INFO) VectorClassifier: standard deviation used: [1, ..., 1]
2019-10-31 16:54:33 (INFO) VectorClassifier: Loading model

and the prediction was complete a few minutes later.

Everything works fine but I'm a bit astonished concerning the time spending to read the input vector : 30 minutes. As far as I understand the VectorClassifier's code, only the function ReadInputListSample is called between this two logs:

2019-10-31 16:23:36 (INFO) VectorClassifier: Model loaded
2019-10-31 16:54:33 (INFO) VectorClassifier: mean used: [0, ..., 0]

Here some details about the input vector (which is in SQLite format):

  • number of samples : 202664
  • number useful fields : 481
  • file's size : 794M

I also used the application TrainVectorClassifier which read exactly the same data. This time the input vector has been read in 12 seconds.

Is it possible read the input vector data in VectorClassifier the same way than TrainVectorClassifier ?

OTB's version : 7.0.0 built in Superbuild mode.