bic not running on mobio

[ankit.gau...@itbhu.ac.in]

NOTE : I know I am using deprecated facereclib instead of bob.bio but I was unable to add mobio to bob.bio.face

Hi,

when I run `./bin/baselines.py --all -vv  --database mobio`

bob.facereclib@2017-03-01 14:59:17,095 -- INFO: Executing algorithm 'bic'
bob.facereclib@2017-03-01 14:59:17,095 -- INFO: Executing command:
./bin/faceverify.py --database mobio --preprocessing face-crop --features grid-graph --tool bic-jets --sub-directory baselines/bic --groups dev eval -vv
bob.facereclib@2017-03-01 14:59:20,194 -- INFO: - Preprocessing: processing 16420 data files from directory '/media/ankit/ceff7ba4-81e0-4dfe-80d7-4fe99720a56d/home/lab/IMAGES_PNG' to directory 'temp/mobio/baselines/bic/preprocessed'
bob.facereclib@2017-03-01 15:09:37,709 -- INFO: - Extraction: extracting 16420 features from directory 'temp/mobio/baselines/bic/preprocessed' to directory 'temp/mobio/baselines/bic/features'
bob.facereclib@2017-03-01 15:19:08,544 -- INFO: - Enrollment: loading 50 enroller training files
bob.facereclib@2017-03-01 15:22:56,528 -- INFO: - Enrollment: training enroller 'temp/mobio/baselines/bic/Enroller.hdf5' using 50 identities:
bob.facereclib@2017-03-01 15:22:56,572 -- INFO:   -> Computing pairs

When I plot the result using `./bin/baselines.py --all -vv --evaluate ROC DET CMC HTER --database mobio`, then I get

bob.db.frgc@2017-03-01 16:16:40,756 -- WARNING: The database directory '[YOUR_FRGC_DIRECTORY]' does not exist. Please choose the correct path, or correct the path in the Interface.frgc_database_directory() function of the bob/db/frgc/driver.py file.
bob.facereclib@2017-03-01 16:16:41,511 -- WARNING: Skipping algorithm 'bic' since the results cannot be found.
bob.facereclib@2017-03-01 16:16:41,511 -- INFO: Executing command:
./bin/evaluate.py --directory results/mobio/baselines --legends eigenface lda gaborgraph lgbphs gmm isv plda --criterion HTER --roc ROC.pdf --det DET.pdf --cmc CMC.pdf -vv --dev-files eigenface/scores/male/nonorm/scores-dev lda/scores/male/nonorm/scores-dev gaborgraph/scores/male/nonorm/scores-dev lgbphs/scores/male/nonorm/scores-dev gmm/scores/male/nonorm/scores-dev isv/scores/male/nonorm/scores-dev plda/scores/male/nonorm/scores-dev --eval-files eigenface/scores/male/nonorm/scores-eval lda/scores/male/nonorm/scores-eval gaborgraph/scores/male/nonorm/scores-eval lgbphs/scores/male/nonorm/scores-eval gmm/scores/male/nonorm/scores-eval isv/scores/male/nonorm/scores-eval plda/scores/male/nonorm/scores-eval
bob.facereclib@2017-03-01 16:16:42,949 -- INFO: Loading 7 score files of the development set
bob.facereclib@2017-03-01 16:16:45,115 -- INFO: Loading 7 score files of the evaluation set
bob.facereclib@2017-03-01 16:16:50,436 -- INFO: Computing HTER on the development and HTER on the evaluation set
The HTER of the development set of 'eigenface' is 37.604%
The HTER of the evaluation set of 'eigenface' is 36.279%
The HTER of the development set of 'lda' is 30.673%
The HTER of the evaluation set of 'lda' is 34.926%
The HTER of the development set of 'gaborgraph' is 8.938%
The HTER of the evaluation set of 'gaborgraph' is 16.790%
The HTER of the development set of 'lgbphs' is 13.348%
The HTER of the evaluation set of 'lgbphs' is 19.192%
The HTER of the development set of 'gmm' is 8.403%
The HTER of the evaluation set of 'gmm' is 11.290%
The HTER of the development set of 'isv' is 3.163%
The HTER of the evaluation set of 'isv' is 7.128%
The HTER of the development set of 'plda' is 21.812%
The HTER of the evaluation set of 'plda' is 23.331%
bob.facereclib@2017-03-01 16:16:50,486 -- INFO: Computing CAR curves on the development and on the evaluation set
bob.facereclib@2017-03-01 16:16:50,582 -- INFO: Plotting ROC curves to file 'ROC.pdf'
bob.facereclib@2017-03-01 16:16:51,889 -- INFO: Computing DET curves on the development and on the evaluation set
bob.facereclib@2017-03-01 16:16:53,261 -- INFO: Plotting DET curves to file 'DET.pdf'
bob.facereclib@2017-03-01 16:16:53,619 -- INFO: Loading CMC data on the development and on the evaluation set
bob.facereclib@2017-03-01 16:16:59,063 -- INFO: Plotting CMC curves to file 'CMC.pdf'

As you can see, no errors are thrown and it ends right after this command. Any idea what am I doing wrong? Why is bic not working while for other db it is working

[Manuel Günther]

Hi Ankit,

it seems to be a bit unusual that there is no error thrown, but I assume that there is a memory issue when computing BIC training pairs. In the BIC configuration file (in `bob.bio.face` though, but the configuration should be called similarly for the facereclib) you can find `maximum_training_pair_count=1000000`, which might be too large to fit into your memory. For some other databases, the number of training pairs is smaller as the limit, so for these databases the execution works fine.

During the evaluation, you can see a line 

bob.facereclib@2017-03-01 16:16:41,511 -- WARNING: Skipping algorithm 'bic' since the results cannot be found.

This indicates that the processing was not successful.

We have just pushed new packages of the bob.bio framework, where we have worked on the integration of databases. Maybe you can give it a try with the new version, too.

Manuel