Speech Emotion Recognition (Sound Classification) | Deep Learning | Python

[@rpvikas]

Hii 👋🏼! I wanted to follow this for recognising baby's cry using their tone of sound . If I used required data set for it will I be able to do it using this.

[@HackersRealm]

yes, it is definitely possible

[@alisher7545]

Thank you so much from my heart I am searching this project from past week some are paid and some didn't share the code you are amazing❤❤❤

[@HackersRealm]

Glad it was helpful!!!

[@Storm_cipher]

Hey how do you get the final output of emotion detected

[@radhebhay]

The video, I was finding the same after searching a lot at last I get it in this video all the content I needed those all contents are mentioned and for what I was searching. I am satisfied with this video. 🥰😇

[@HackersRealm]

Glad it was helpful!!!

[@laurapagani321]

Thank you for your code. Could you please add how can I test the model with a new audio file?

[@yikes3807]

Hey did You found out? can you help?

[@varunkareti8628]

Hey did you got the code

[@HackersRealm]

Hi Everyone!!! New video with transformers has been released (99.6% accuracy) and link is available in description with latest code. For the current video, due to some hidden files in kaggle data, the dataset has been loaded 2 times. Please check the github repo for the updated code file. The validation accuracy will be around 67% which is better for a baseline multiclassification. The process remains the same for the project. Due to module updates, there are few changes needs to be done. It has been updated in github notebook. 1. sns.countplot(data=df, x='label') 2. In waveplot function, update- librosa.display.waveshow(data, sr=sr). Use waveshow instead of waveplot in librosa Happy Learning!!!!

[@devisnugroho]

how to load dataset if I had in my local directory?

[@HackersRealm]

@@devisnugroho You can just set the path in base_dir variable in a similar way

[@rittataylor_2000]

@@HackersRealm sir can you share the papers you used for reference in this topic

[@HackersRealm]

@@rittataylor_2000 Sorry I didn't use any papers for reference

[@rittataylor_2000]

@@HackersRealm sir can you share description of the model that you have used for this project

[@sagarbhagwani7193]

Sir amazing project with beautiful explaination just one question why LSTM ? we all know they work good with textual data

[@HackersRealm]

Currently it's better to go with transformers as it will work much better than LSTM

[@sagarbhagwani7193]

i agree sir but i was confused that why lstm Transformers has that capability to work with different type of datas . obviously it will also perform best withbtest but it can work with images and tabuler data too so why lstm? @@HackersRealm

[@HackersRealm]

@@sagarbhagwani7193 As i said, it was just for tutorial purpose. New video will be made soon in the upcoming months.

[@bakhshizade]

Thanks for this tutorial. I am interested in implementing personality trait speech recognition, which I think is somehow related to emotion recognition. Any idea/advice on this topic? Thank you!

[@DivyaPrakashMishra1810]

Thanks for the detailed explanation, just some questions as I'm new to this. So the models are using features like mfcc for recognition, that means the waveplot and spectrogram is for visual representation only. right?, Also how can I make changes so that I give the audio file as input and the model gives emotion as output?

[@HackersRealm]

waveplot and spectogram is for visualizations only... and for the model prediction, you can do the similar feature extraction steps and use model.predict to extract features.

[@sumantralal1029]

bro did you made this project??

[@vrnkareti7154]

did you got the output by giving an audio file

[@Storm_cipher]

Hey brother, how do i get the final output, that is the final emotional classification eg happy sad anger etc????? Please tell!!!

[@PrinceGhotar]

Thanks you so much for your tutorial but can you please tell how this helps to extract stress score.

[@HackersRealm]

I think we can't use this for stress score estimation

[@PrinceGhotar]

@@HackersRealm okay, thanks for your reply.

[@id486-wilgenjacobo8]

hey brother, excellent video, currently I run it in Jupyter notebook and if it turned out very well, I need some advice, if I want to try it with a different audio, how can I do it?

[@HackersRealm]

You have to use the same preprocessing steps and use model.predict to predict the class

[@LamNguyen-hw9lq]

What are the train and test set in this case?

[@P.S.D_99]

hi bro ...your project is really helpful but apart from accuracy does it classify which emotion is depicted in the audio file?

[@HackersRealm]

yes, that's the purpose of project. I plan to make video with latest model which can do it realtime. Please wait for sometime for that.

[@Storm_cipher]

Hey brother do you know how to get the output if yes please reply

[@LitalLevy-m2q]

Hello, thanks for the tutorial! Can the model be used for real-time emotion recognition of a random voice?

[@HackersRealm]

yes, you can

[@eminyazilim]

Thanks for the video, but after the model is trained at the end, which label is the trained model? feat,angry vs. This stage is important for me, would you be interested?

[@HackersRealm]

you can use that one hot encoder inverse transform to get the original values (or) you can also see the class labels using that enc object

[@benedictespiritu1001]

Am I correct to assume that because the training acc and validation acc are not far from one another then the model is not overfitting and performing well? If the validation acc is less than the training acc then it is overfitting, right? Thank you for your answers.

[@HackersRealm]

Yes you're right. If the val accuracy is not improving and the train accuracy is increasing then it's a sign of overfitting

[@benedictespiritu1001]

@@HackersRealm may I ask why the number of units for the LSTM was set to 128? Thank you.

[@ScottRock12]

Thank u, what a tutorial!

[@pavankumar-ue9fd]

bro i have doubt in loading of data set can u explain in brief ??

[@HackersRealm]

yeah please share the query here

[@sarfarazahmed9614]

Please make a video on how to deploy this model on a web application using flask. That will be great!

[@HackersRealm]

Well, I already made a video how to deploy any model using flask, kindly check it

[@daviddejesusgallegospoblan2068]

Hi, how can I predict which category a new audio belongs to? Excellent video

[@HackersRealm]

you have to do the same preprocessing what we have done for the dataset and predict it.

[@rittataylor_2000]

@@HackersRealm sir in the dataset when we do features extraction is it stored and labelled as these features belong to a specific emotion using the os files and use the classified dataset to teach the model which detect which mfcc features belong to which emotion

[@MSRRAJPUT-yx6ku]

Hello sir the explanation was awesome but i a getting lot of error can i get some help

[@preethikae8519]

Was there a reason why you chose LSTM-RNN model because MLP classifier gives more accuracy

[@HackersRealm]

I just choose to show it as an example. To find the best model, we can try out few more different models to finalize

[@adnanshahid3452]

Hello SIr I hope you are fine in this code you havent elaborated that how to test this model like how will give file input in the form of my voice and model will tell me that am I angry happy or disgust?

[@akshatashadakshari5381]

Hi sir ,my final project topic is " To predict the voices of siblings are matches"..is it possible to do that?

[@HackersRealm]

Yes, if you have the proper dataset, you can use the same workflow for that

[@sreekrishna8]

Bro , what are these 40 features ?

[@HackersRealm]

40 is just the no. of reduced features

[@tetakalapavani7510]

will this work in jupyter too?

[@HackersRealm]

yes

[@adhamebaid7758]

thanks for the vedio , how can i predict a single audio file after training the model?

[@sumantralal1029]

did you made this project with any frontend

[@Storm_cipher]

Did you do it???

[@mouskey]

Brother can you upload the video about gesture volume control ?

[@HackersRealm]

I am planning to do gesture recognition project in few weeks for sure!!!

[@mouskey]

@@HackersRealm tq sir

[@Raysworkshop]

Hi! I've followed the tutorial from top to bottom. If you do it with the same way in the tutorial, you get a close to 100% val_accuracy, but that's only because the training image and the testing image are the same. If you do it with half the image, like in the Github files, you get a around 25% val_accuracy, which is a lot more realistic but really ineffective. Is there any way to improve the val_accuracy without "cheating" and using the same data for testing and training?

[@HackersRealm]

If you check the github, the val accuracy I have got is around 67, which is better. the github currently not showing the output results from last month. That's why I updated the readme with the accuracy I have got. You can improve by using additional features, changing network and tuning it.

[@Raed_Matti]

Do you have a video explains how to make lstm with k-fold cross validation?!

[@HackersRealm]

for deep learning, I don't have that code

[@muscity36]

have u published any paper regarding this project?

[@HackersRealm]

No

[@sowmyareddy6866]

in exploratory data analysis im getting what should I do

[@HackersRealm]

Are you facing any errors, if yes, what is it?

[@laiqamalik8345]

Can we do this project with CNN or RNN?

[@HackersRealm]

yeah, it's possible

[@thaslina3470]

where to run and get the output?

[@arryansinha1158]

can you tell how to deploy this project

[@HackersRealm]

you can watch this video to get an idea on how to deploy any ml project kzbin.info/www/bejne/aH3Uo5l7r9p8g80

[@sreyamathew327]

How to do ser using CNN? I'm a beginner.

[@awatefmessaoudi5666]

thank you for this video, could you please help me to convert the discrete emotion to valence arousal and dominance

[@HackersRealm]

I Couldn't get the last part

[@pushpacarpenter9522]

Training data contains 1 samples, which is not sufficient to split it into a validation and training set as specified by `validation_split=0.1`. Either provide more data, or a different value for the `validation_split` argument. I am facing this error while training of model, how can i solve this? and i am using kaggle colaboratory

[@HackersRealm]

I think there is a problem in dataset and some class have less samples, please check that once as it might be causing the issues.

[@hakimarab2436]

i did the same steps, but i'm having 400 audio per label, but not 800 ?, this will affect the training

[@HackersRealm]

in the video, the samples were repeated twice, that's why... 400 is the correct flow... I already pinned the comment on that

[@jayasreeskitchen6220]

Bro can you help me out in How to predict the emotion with the model trained

[@HackersRealm]

Could you share which specific part in predicting the class you need help. In general, you have preprocess the data as before and use model.predict() to get the output

[@chittisai47]

Hey! Thank you so much. That helped a a lot. I have a question can I do the same using train_test_split ??

[@HackersRealm]

yeah, you can use it

[@chittisai47]

@@HackersRealm What is the difference between validation split and train_test_split

[@HackersRealm]

@@chittisai47 test split and val split are same when we are considering in this aspect

[@chittisai47]

@@HackersRealm Ok. Thank you!!

[@chittisai47]

@@HackersRealm Sir can you please tell me how do I get the accuracy percentage of the model. I want to print it on the model at last.. What should I do?

[@aesthetic_shobiz_life]

how i show the accuracy on test dataset ?

[@sakthiyajayakumar8807]

Hey there, I'm getting an ValueError: could not convert string to float:'fear'. In the line: sns.countplot(df['label']) what could be the issue? Thanks

[@roojina9525]

Hi I have the same error. @HackersRealm Would you please help?

[@sakthiyajayakumar8807]

@@roojina9525 Hi, yea he seems to replying to new comments too similar to ours, but still haven't replied to ours. Let's see

[@asgaraizazpeerzada8512]

any1 solved this error

[@KaranBhanushali-gt6ts]

@@asgaraizazpeerzada8512 use this -- > sns.countplot(x=df["label"])

[@dhruvgoyal4261]

sir can you please provide syntax for accuracy_score

[@Aastha-x4o]

How can i choose a particular audio file in the given code?

[@allaveerabadrikumar7625]

Hi sir my major project is "automated notes maker from audio recordings" It is just speech to text convertion. How can I do it what is the algorithm I have to use can you please help me

[@HackersRealm]

Check out image caption generator video, it's similar to your project flow. Instead of image, you have to use audio as input

[@o__bean__o]

Bro you lucky 😂 you got Training accuracy 99% Validation Accuracy 99% in first Start

[@HackersRealm]

Kindly check the new video which was released recently with transformers.

[@moviehacks7866]

Im unable to load the dataset in jupyter,can you plss explain?

[@HackersRealm]

I was using kaggle notebook, please use the correct path where you have placed your dataset

[@chittisai47]

Sir you said your updated accuracy is 67% in github. How did you say it?? Please let me know. Can we print the accuracy of the model in last of the model If so please let me know how to do it..

[@HackersRealm]

if you use checkpoint and save the model based on best val accuracy, you can get the mentioned result model

[@roshanaditya8739]

Can we save this model of the form '.h5' and use it for another model? If so how?

[@HackersRealm]

yes you could save it as h5 but when you load, it will be same model architecture and weights

[@laurapagani321]

I did a prediction but the model returns always the same class: how can it be?

[@HackersRealm]

If it's always giving same class for any input, then the model might have not trained well or it may be due to some error in doing the prediction

[@jitinchand4460]

Sir I am getting an error showing that melspectrogram() takes 0 positional arguments but 1 positional argument (and 1 keyword-only argument) were given.

[@preethikae8519]

Make this change in the extract_feature function mel=np.mean(librosa.feature.melspectrogram(y=X, sr=sample_rate).T,axis=0)

[@lifetube1117]

Thanks very much !

[@HackersRealm]

Happy to help!!!

[@nivy977]

Bro can you teach me how this code work in python i facing lot to trouble

[@HackersRealm]

could you share where you're facing troubles?

[@prathameshyadav6090]

AttributeError Traceback (most recent call last) Cell In[52], line 4 2 path=df['speech'][df['label']==emotion][0] 3 data,sampling_rate=librosa.load(path) ----> 4 waveplot(data,sampling_rate,emotion) 5 spectogram(data,sampling_rate,emotion) Cell In[51], line 4, in waveplot(data, sr, emotion) 2 plt.figure(figsize=(10,4)) 3 plt.title(emotion,size=20) ----> 4 librosa.display.waveplot(data, sr=sr) 5 plt.show() AttributeError: module 'librosa.display' has no attribute 'waveplot'

[@HackersRealm]

instead of waveplot, use waveshow()

[@smarthanand4327]

Sir is there any GUI related this ?

[@HackersRealm]

sorry, I don't have any GUI for this one

[@Haritesh-p9z]

@@HackersRealm then make one

[@Mickymouse340]

Hey, good video. I just wanted help with using the inverse_transform() function on the onehotencoder. I have the model and want to use it to predict a new audio file. I'm not sure what the parameters of the enc.inverse_transform() would be.

[@HackersRealm]

In the inverse transform, you have to pass the predicted output from the model. So it would convert the integers back to string classes

[@rittataylor_2000]

@@HackersRealm and when we create the lstm layers do we set the layers in such a way that the weight and biases of the model based on the input features will activate the neuron in the output layer that has the highest possibility of the emotion conveyed by the speech

[@praveenkumawat8183]

How to create setup for this project

[@HackersRealm]

You can directly go to kaggle and use this notebook, it's simple only

[@jatinsadhwani9218]

Sir please deploy it using streamlit or flask✨✨.. It will be useful for us

[@HackersRealm]

yeah sure, model deployment using flask video will be uploading soon!!!

[@swagsterfut9992]

@@HackersRealm actively waiting, please upload it soon Sir

[@HackersRealm]

@@swagsterfut9992 already uploaded kzbin.info/www/bejne/aH3Uo5l7r9p8g80

[@swagsterfut9992]

@@HackersRealm woah thanks, does code change much when deploying an audio model. also if you have deployed the above project through flask/django, that would really help a lot!

[@HackersRealm]

@@swagsterfut9992 The flow remains the same for all the projects, only thing is you may need to add preprocessing step. No, I didn't have code for the above project

[@urjaalwadkar304]

How to standardise our dataset here? Pls help

[@sam-mv6vj]

thanks bro

[@HackersRealm]

Happy to help!!!

[@AddoDankwahBaffour]

sir with the data analysis am having error with that says "could not convert string to float: 'fear'"

[@HackersRealm]

I have checked now, it's working fine!!!

[@anaahatulhe223]

Sir, I am getting an error like this-"NotImplementedError: Cannot convert a symbolic Tensor (lstm_3/strided_slice:0) to a numpy array. This error may indicate that you're trying to pass a Tensor to a NumPy call, which is not supported" please help thanks

[@HackersRealm]

Did you change the implementation from the notebook? Which part you're getting this error?

[@meenakrish6875]

bro one doubt after executing the loading the dataset cell if i execute paths or labels it is not shwoing the paths or labels like in ur video what to do for that

[@HackersRealm]

Are you using the same dataset or different one?

[@Stains-lz9pz]

Bro I got error in the final part... history = model.fit(X, y.toarray(), validation_split=0.2, epochs=100, batch_size = 512, shuffle = True) ValueError: Failed to convert a NumPy array to a Tensor (Unsupported object type numpy.ndarray). what to do?????

[@HackersRealm]

Can you check whether both x and y are numpy array?

[@aandabros491]

How can I detect an emotion when giving a voice input.

[@HackersRealm]

You can capture the audio and use the same preprocessing methods and predict the label from the model

[@aandabros491]

Is the code same

[@aandabros491]

Is there any changes

[@HackersRealm]

@@aandabros491 You have to make changes for getting real time input!!!

[@kanchansahitani4857]

emotion = 'fear' path = np.array(df['speech'][df['label']==emotion])[0] data, sampling_rate = librosa.load(path) waveplot(data, sampling_rate, emotion) spectogram(data, sampling_rate, emotion) Audio(path) I am getting error there AttributeError: '_process_plot_var_args' object has no attribute 'prop_cycler' please help me

[@Haritesh-p9z]

Try/ pip install scipy

[@btr_fun_gamerz1641]

Value error in the exploratory data analysis part

[@kinaririos254]

same

[@dexnug]

hi I've tried to run with only 5 classes data, everything work fine, except I my validation acc is zero, and val loss getting higher..can you please help me to fix this?thanks

[@HackersRealm]

the workflow remains the same for any number of classes, can you reach me via insta or linkedin

[@prajwal3117]

@@HackersRealm i have the similar issue can you share your instagram or linkedin Profile ID

[@sam-mv6vj]

Bro have you uploaded breast cancer classification yet ??

[@HackersRealm]

Not yet, will be creating video with different techniques as I covered so many classifications problems till now!!!

[@sam-mv6vj]

@@HackersRealm i mean i have seen in commuinity tab list of videos that you gonna upload and my project also about breast cancer detection that why i asked

[@HackersRealm]

@@sam-mv6vj okk will be uploading soon!!!

[@sam-mv6vj]

@@HackersRealm thank you so much

[@advisionmegaconcept6128]

please how do i run the file directory path on jupyter notebook?

[@HackersRealm]

change this os.walk('/kaggle/input') to this os.walk('') in data loading

[@advisionmegaconcept6128]

Thanks for your prompt response. This is my challenge, I kept having empty cell in paths[] and labels[] which I knew because I have not gotten loading data code part right on jupyter notebook. Help out please .

[@HackersRealm]

@@advisionmegaconcept6128 for loading the data only, i have mentioned in the above reply

[@advisionmegaconcept6128]

Yea but it loads an empty dataset.

[@advisionmegaconcept6128]

@@HackersRealm thanks, all sorted out and running.

[@nimishamr3940]

hey im getting a error in exploratory data analysis catogorical data error,apparently i should one hot encode,what do i do

[@HackersRealm]

Could you tell me where exactly you're facing this error? Are you using the same notebook and dataset?

[@asgaraizazpeerzada8512]

u solved it if yes plz help

[@dexnug]

how if we want to check performance of model with the new audio data file?

[@HackersRealm]

you can preprocess the data in same way and use model.predict to make a prediction

[@Storm_cipher]

​@@HackersRealmhow to do this thing bro please reply

[@Storm_cipher]

Bro did you do it

[@archanapanjada3151]

How to clear the waveplot error?

[@HackersRealm]

you could use waveshow or use different library

[@sam-mv6vj]

Bro do few end to end project also

[@HackersRealm]

end to end project means? Most of the projects are covered all the topics... Can you explain briefly

[@foods-travel]

Test File is necessary plzzzz

[@HackersRealm]

you can split the data for train and test. There is no separate test file for this dataset

[@harithak7114]

Hai I am getting error that module librosa has no attribute to display Please help me

[@HackersRealm]

Try to force reinstall the library using pip

[@mousikaa8429]

bro how can i access this dataset when i run it in collab or other compilers

[@HackersRealm]

you can download the data from kaggle, link in the description

[@maheswari8691]

I got error at feature extraction it is name df is not defined.can you please solve this

[@HackersRealm]

try reruning the notebook again, some cells you may be missed while running

[@maheswari8691]

@@HackersRealm I got another error it is y.toarray() it shows numpy.ndarray object has no attribute

[@meghana4037]

Hey, I cleared with the waveplot error…now in feature extraction, X_mfcc=df[‘speech’].apply(lambda x: extract mfcc(x)). This is taking long time to run the code and I’m getting timeout zone.

[@HackersRealm]

yes this will take a long time to extract features

[@meeramohan9521]

I got value error while train the model Shapes (none, 8) and (none, 7) are incompatible. How can i resolve that

[@HackersRealm]

Did you use the same the notebook?

[@chittisai47]

I got the same. But I did some input error. Please Check..

[@alok_ranjan5163]

i am getting error while train the model

[@HackersRealm]

in which part you're getting error?

[@alok_ranjan5163]

Train the model, but i solved that error.​@@HackersRealm

[@nimishamrelectronicsengine4453]

In feature extraction of mfcc,y=y.toarray isnt working because such an attribute doesnt exist itseems.What do i do,

[@sushmithagajalakshmi153]

Have you rectified the error?

[@supravatswain]

bro, history = model.fit(X, y, validation_split = 0.2, epochs = 50, batch_size = 128) if I train the model using this i got the error ValueError: Shapes (None, 1) and (None, 7) are incompatible. I'm using LabelEncoder for it. plz help me here.

[@HackersRealm]

you may have to change part of the model if you're changing to label encoder

[@supravatswain]

@@HackersRealm bro I changed the model part but getting same error.

[@foods-travel]

hello sir plzzzz test file share

[@meghanaarballi7270]

Hey, I am getting an Attribute Error for Emotion=fear

[@HackersRealm]

Are you using same dataset?

[@meghanaarballi7270]

Yess same dataset

[@HackersRealm]

@@meghanaarballi7270 If you're using same code, you won't get the issues like this or you can reach me via insta or linkedin to share the pic of the error

[@meghanaarballi7270]

We have used the same code but still getting this error

[@sahilbhatt2181]

sir how can i give another audio to model and predict the label ? can you please give me code snippet ? please !!

[@HackersRealm]

you have to do the same preprocessing steps we have done for the audio and do the prediction from the trained model

[@sahilbhatt2181]

def feature(filename): y, sr = librosa.load(filename, duration=3, offset=0.5) mfcc = np.mean(librosa.feature.mfcc(y=y, sr=sr, n_mfcc=40).T, axis=0) X = [x for x in mfcc] X = np.array(X) X = np.expand_dims(X, -1) return X ftr = feature(r'E:\SAM\SEM 6 PROJECT\Recording.wav') prediction = model.predict(ftr) print(prediction.shape) i've done this but yet don't know how can i retrieve label can you please guide me ? please !!

[@HackersRealm]

@@sahilbhatt2181 you want to save this on the disk?

[@sahilbhatt2181]

@@HackersRealm yes sir ! I'm coding it in vs code.

[@HackersRealm]

@@sahilbhatt2181 if you are predicting for multiple files, use list to save the predictions and create a csv from that. It's easier that way

[@Arjun-ks1gh]

getting error in sns.countplot(df['label']) step please help

[@HackersRealm]

have you tried running from start?

[@Arjun-ks1gh]

@@HackersRealm bro only that line is not working rest is fine. Anyways its working..Thank you..

[@asgaraizazpeerzada8512]

@@Arjun-ks1gh bro how u solved that error plzz help

[@gundamdinesh5067]

How to run the code ?

[@HackersRealm]

you can run in google colab or jupyter notebook

[@shehabeldin8307]

hello i wanna somebody to help me when i using waveplot it give me error i wanna to know whats wrong ??!!

[@HackersRealm]

what error it's showing?

[@shehabeldin8307]

@@HackersRealm AttributeError: module 'librosa.display' has no attribute 'waveplot'

[@HackersRealm]

@@shehabeldin8307 did you install librosa module? you can do using "pip install librosa"

[@shehabeldin8307]

@@HackersRealm yes yes i used pip install librosa

[@HackersRealm]

@@shehabeldin8307 still facing the issue? are you running in local or kaggle?

[@foods-travel]

Make it plzzz

[@dilipprince9181]

if i type the coding as below emotion='fear' path = df['speech'][df['label']==emotion][0] data, sampling_rate = librosa.load(path) waveplot(data, sampling_rate, emotion) spectogram(data, sampling_rate, emotion) AttributeError: module 'librosa.display' has no attribute 'waveplot' i got attribute error what i have to do

[@HackersRealm]

It depends on the module version, i think some comments have alternate solution

[@yolandasundari5356]

@@HackersRealm hi sir, i want to ask, that means in this video, you use algorithm "RNN" right? or any else? hope your answer, thanks

[@HackersRealm]

@@yolandasundari5356 yes it rnn architecture only

[@siddharthsingh8775]

I might be a bit late to this but instead of waveplot it should be waveshow().

[@suhbhajeetkrishnadey7415]

Bro there is a problem in Audio(path) Name ‘audio’ is not defined Idk why🥲

[@HackersRealm]

you have to import that audio module i guess

[@dhruvgoyal4261]

pls help the sns.countplot(df['label']) was not working and showing errors pls help

[@HackersRealm]

Use sns.countplot(x='label', data=df)

[@dhruvgoyal4261]

@@HackersRealm thanks 🙏