vincent-zhq/ca-mser

Code for Speech Emotion Recognition with Co-Attention based Multi-level Acoustic Information

Hi there 👋

ca-mser's Issues

extract_features.py

Dear author, I am conducting the SER project, and your project is helpful to me. But I found no extract_features.py file mentioned in your project. Could you please share the code? My email address is [email protected]. Thank you very much.

about the extract_features code

你好，请问可以分享一下文件预处理的代码吗？即extract_features.py。非常感谢!

Could you please send me data preprocessing and feature extraction codes?My email address is [email protected]. Thank you very much！

Data preprocessing

Hello, thanks for your great project! Could you please provide the code how to preprocess data from .wav file. I wonder how to make spectrogram with dimension (300*200), which was mentioned in the paper.

Feature Extraction

Could you please send me the source code for data preprocessing and feature extraction, i.e. extract_features.py . My email is [email protected].
Thanks a lot!

May I ask how to use it？

How can I correct the following error?
FileNotFoundError: [Errno 2] No such file or directory: 'IEMOCAP_multi.pkl'**

About train_loss = train_loss_ce + train_loss_mml

Hi, I hava a question. I just want to see the effect when I use train_loss_mml in your code.
like this:
train_loss_mml = criterion_mml(outputs['M'], train_labels_batch)
train_loss = train_loss_ce + train_loss_mml
and the following warning information is generated
[W Resize.cpp:23] Warning: An output with one or more elements was resized since it had shape [], which does not match the required output shape [32, 4].This behavior is deprecated, and in a future PyTorch release outputs will not be resized unless they have zero elements. You can explicitly reuse an out tensor t by resizing it, inplace, to zero elements with t.resize_(0). (function resize_output_check)

RuntimeError: Input type (torch.cuda.FloatTensor) and weight type (torch.FloatTensor) should be the same

excuse,can you tell me about above error? why?

System RAM Crashes

Hi sir,
Good time,
Thank you very much for the codes of your article,
The codes are written very neatly and readably,
I run these codes in the Colab environment, but I have the only problem is that Colab 's System RAM crashes,
(In Colab i have:
System RAM = 12.7 G
GPU RAM = 15 G )

What is the amount System RAM you run this program on?
How much system RAM do I need?

problem solved

May I ask if the total number of iterations is 200？200= repeat_kfold * len(val_id) * len(test_id)

I am a beginner and I think your paper and code are very ingenious. However, I have one more question and hope you can answer it.Thanks！
1What does True mean in the output result？2What does 31 31 mean？
True
31 31

accuracy values for the training

Hello, I printed the validation set accuracy and it is showing normally at around 71. However, the two accuracy values for the training set are both around 25, and I'm not sure why that is happening.

About the results

Hello, excuse me, are the average UA and WA results of the source code running the results given in the paper?

About IEMOCAP Accuracy

Hi, I have a question. Obviously I know that the IEMOCAP dataset is multi-label, but if you look at the code, it seems that the accuracy was measured with single-label method. Maybe I misunderstood the code?

If you don't mind, can you explain it to me?

Thank you

run question

Hello，When I run these lines of codes
for i, seg in enumerate(data):
seg = np.clip(seg, 0.0, 1.0)
seg_rgb = (cm(seg)[:,:,:3]*255.0).astype(np.uint8)

        img = Image.fromarray(seg_rgb, mode='RGB')

        data_tensor.append(alexnet_preprocess(img))
   
    return data_tensor

---

Dataset normalized with minmax scaler
Range before normalization: [-80.0, 3.8146973e-06]
Range after normalization: [0.0, 1.0]
已杀死

The above error occurs

Best wish to you！

About data processing code

Hello, I see you say update data processing code. I didn't find it.

quesition

Hello, I want to ask what dataset IEMOCAP_multi.pkl is? Is it convenient to share it?

How to inference for a wav

How to inference for a wav ?

Some questions for Speech Emotion Recognition

Dear Professor,
I am a master of Hebei University of Science and Technology,major in Computer Science and Technology.Your published paper,"Speech Emotion Recogition with Co-Attention based Multi-level Acoustic Information", gave me lots of inspirations,especially the feature extraction part and the feature fusion part.But I noticed that there is a part of source code had disappeared,named "feature_extraction".if it is convenient for you,could you please sent it to my email,[email protected],I believe it will help me a lot.
Best Wishes,
Jialin

Question about the extract_features code

Could you please send me the source code for data preprocessing and feature extraction, i.e. extract_features.py . My email is [email protected]. Thanks a lot!