Valeriu Lacatusu

Facebook moved to a different S3 bucket See: https://github.com/facebookresearch/InferSent/issues/93

Question

4

Why do we need to do this in fact: "Use ffmpeg to convert and split WAV files into 10 second parts"? After downloading we have big wav files. We can...

Training convergence

8

I'm trying to train the crnn network using latest keras on 2 languages (English and French) on a "youtube spoken" dataset. But it seems the validation accuracy (and not only)...

Pull Request

13

I've included all your different advancements and added new functionality since the last time : 1° Added the InsertBatchWithIdentity ( which uses oracle bulk options with OdpDataProvider and Batch insertion...

msr_sampler is needed only when coco == True

New issues

1

1° Reading CSV files without separator char (Add CharLength property in field csv attribute) 2° Using FieldIndex for reading data at the right index in the data row 3° Using...

For some images (mainly the ones taken with a phone) the library inverts height and width dimensions. For the attached image (which is a landscape image of a document) -...

Training details

1

What loss, accuracy you obtain when training with the compact_cnn model? I'm using the default parameters and loading model like this: model = main('tagger'). But I get a huge overfit...

Is it worth training on MSD with some of your other models from the models_time_invariant script? Ex: - model_crnn_icassp2017_choi, - model_conv3x3_ismir2016_choi, - model_multi_kernel_shape, - model_conv1d_icassp2014_sander.