jefflai108 / attentive-filtering-network Goto Github PK

When I try to run the code using ./run_feature.sh, I get the following error:

./run_feature.sh: line 5: ./cmd.sh: No such file or directory
./run_feature.sh: line 6: ./path.sh: No such file or directory
./run_feature.sh: line 21: utils/fix_data_dir.sh: No such file or directory

Where do I get these files from?
Kindly update the instructions in the README file accordingly.

For run_feature.sh, no such file data/*_cqcc_spectrogram/utt2spk, can you help to upload the data directory?
For run.sh, no module named 'adv_kaldi_io'.
For make_spectrogram.sh, line 18: parse_options.sh: No such file or directory.
Can u help to upload these files? Thank u very much!

there are several questions when I run the project

1. when I run run_feature.sh according to the readme, it occurs the errors about no train_cqcc_spectrogram；
2. should I change the stage in run_feature.sh, change -4 to 4 in line 53 in run_feature.sh;
3. i cannot generate cqcc feature using kaldi

Hi,
when I run the code, I find that there are 9 dataset.py, that is v1_dataset.py ~ v9_dataset.py，
In the git code, main.py use v3_dataset.py to read the train set,
but v3_dataset.py has no code to reshape feature to [1091, 257]。
please give some guide to use which dataset.py

Hi Jeff,
I cannot conclude how you generate the utt2systemID files:

• 1. features/run_feature.sh
     Generates the scp+ark files as usual.
• 2. assert/data_reader/convertID2index.py
     Both functions convert_la() + convert_pa() generate the utt2index files.
     The scp files where generated by run_feature.sh at step i.
     But where do the utt2systemID files come from, how to you generate them, what do they contain?
     Option a) You directly use the utt2systemID files contained in PA/ASVspoof2019_PA_asv_protocols or PA/ASVspoof2019_PA_cm_protocols?
     Option b) You have a script which parses these ASV/CM protocols into utterance ID and system ID?
• 3. data_files['eval_utt2systemID'] is not defined in my_config().data_files{}
     After training with work() you call forward_pass() to conduct the evaluation.
     At the end of forward_pass() you call
     eval_prediction(eval_loader, model, device, score_file_pth, data_files['eval_utt2systemID'], use_rnn, focal_obj)
     But 'eval_utt2systemID' is nowhere defined (just to tell).

Kind regards,

I can't find any information about .scp files.
Can you provide more details about it? Or how to prepare data to train this model?
Thank you in advance.