Part 2-Data Preprocessing-Audio Classification Project Using Deep Learning

[@fenixchow1]

Hi Krish, thank you so much for all these great lectures, really really appriciante!!! You contribution to data science community is invaluable!

[@pankajshaw674]

What a treat! 2 videos in one day.

[@maazirfan]

Thank you so much for teaching us, I got 90% accuracy with KNN, 72% with decision tree, 52% with logistic regression

[@tonyyu5]

May I ask how u implemented KNN with this neural net?

[@maazirfan]

@@tonyyu5 #donot use if using ml algo #from tensorflow.keras.utils import to_categorical #from sklearn.preprocessing import LabelEncoder #labelencoder=LabelEncoder() #y=to_categorical(labelencoder.fit_transform(y))X=np.array(extracted_features_df['feature'].tolist()) y=np.array(extracted_features_df['class'].tolist()) X_train,X_test,y_train,y_test=train_test_split(X,y,test_size=0.1,random_state=0)

[@tlikesh960]

THANK YOU so much for this playlist it really helps me with my mini project( music genre classification ), and your explanation made it much more simple.

[@GlobalApex]

Great explanation. Love from Pakistan 😍. Subscribed

[@abhishek_maity]

I guess before starting a new playlist it would take same time to upload the remaining 1 video for the tfod custom object detection ..( which is the last and main part of that playlist ) So that playlists will not remain incomplete ....... Anyways this was just a suggestion !! Good work keep going 🙂🤘

[@asiftandel8750]

Great video and explaination Sir

[@himangshukakati313]

Can anyone help me with this? ModuleNotFoundError: No module named 'resampy' This error is lazily reported, having originally occured in File /usr/local/lib/python3.10/dist-packages/librosa/core/audio.py, line 32, in ----> resampy = lazy.load("resampy")

[@annmariyafrancis5630]

Could you solve this error

[@arunkanagolkar6006]

install resampy using 'pip install resampy' and Rerun kernal

[@akashchoudhuri3932]

Hello Sir, please upload more ML and DL end to end Project videos along with deployment.

[@jatinsadhwani9218]

Sur please make a detailed practical video on Bert classification model

[@hadjdaoudmomo9534]

Wonderful explanation, thank you so much.

[@sahilkadu9679]

What is the reason behind this line? mfccs_scaled_features = np.mean(mfccs_features.T,axis=0)

[@ompatil6495]

Hey krish I have used this code for my personal project for birdcall recognition but it keep showing resampy error import numpy as np from tqdm import tqdm ### Now we iterate through every audio file and extract features ### using Mel-Frequency Cepstral Coefficients extracted_features=[] for index_num,row in tqdm(metadata.iterrows()): file_name = os.path.join(os.path.abspath(audio_dataset_path),str(row["ebird_code"])+'/',str(row["filename"])) final_class_labels=row["species"] data=features_extractor(file_name) extracted_features.append([data,final_class_labels]) can nay one help me

[@bolopandeji2865]

same here, have been getting resampy error ModuleNotFoundError: No module named 'resampy' This error is lazily reported, having originally occured in File E:\college\anaconda\Lib\site-packages\librosa\core\audio.py, line 33, in ----> resampy = lazy.load("resampy") I have even tried to install resampy again

[@ompatil6495]

@@bolopandeji2865 I sorted my error it was due to dataset having two types of audio files mono and stereo remove all stereo files hopefully it'll sort your error

[@bavithagurram8262]

​@@ompatil6495 can you share the updated code?

[@gregoiremafutala9897]

Hello, thank you for this great video very well explained, thank you! 10:05 I just didn't understand why did you average the transpose of this tensor via np.mean(mfccs.T, axis=0) ?

[@Sibixpur]

samee Doubtt!! , why did he transpose it??

[@shaurya9055]

First of all you are doing an amazing work I love watching your videos but I just want to say one thing Whenever you make any video please use conditions as well in order to explain something because we all have different things and that's why we get stuck so please do not rush the things up Even if the video gets longer no problem if we need it we'll watch it Actually I'm a noob in programming and I just want you to see the things from our perspective. Thanking you!

[@patanjalmishra8740]

Sir please upload a video which shows difference between k-mean and Knn

[@naveenkavindu6800]

Sir please upload Real-time Audio classification

[@OMPRAKASHKUSHWAHA-k8l]

hello sir, i want to extract mfcc feature form data set and apply clustering techniques, how we can do this??( i have doubt regarding how we can extract mfcc features for unlabelled dataset .

[@sayantikachakraborty2055]

Sir the dataset that i am working on doesnt have a csv file and just has the audio..How do i go ahead without having any csv file data?

[@harshitachouhan7654]

Did you got the solution?

[@rifff5]

Same here!

[@032lovishkumar8]

for resampy error , try installing resampy , importing it and then restarting kernel

[@shanmukhasreevatsa728]

Thank you so much for the wonderful video... certainly helped me a lot! I just have one quick question. In the feature_extractor function you defined, what is the purpose of taking the mean for MFCC values? i.e. why is mfccs_scaled_features = np.mean(mfccs.T, axis=0) useful? What does it actually do, please help me out. Thank you!

[@ryankuruppu7297]

Hi, I was just following the same thing. If we don't use it it seems to give out a nested array. When we use the np.mean function it uses an array of the means of the nested arrays.

[@durgaganesh423]

Hi thank you for this video I have one query how to record loopback from default endpoints ? Any solution And is it possible find glitches or noise from loopback audio

[@jasimali3981]

you explain really good , please help me with i want to train my dataset i.e i have class folders and wav files inside it but i dont have csv file , how can i train my dataset ?

[@abdulmeral4811]

why did you do encoding for labels? do we need to do that? thank you

[@tonycardinal413]

Aweseome video! Quick ques: How do I convert a wave file to a 128x128 spectogram?

[@rakeshdeo6197]

Thank you, what is the total course duration of all these.

[@samansaeedi102]

dear Krish the dataset seems to be unavailable for download. I also tried to download it using soundata library but there is still error. can you please upload the data set somewhere that those who need it can download it?

[@pipicovers]

awesome video. i am getting very less accuracy around 36% for my target variable. is there any way to improve it ?

[@souravmohapatra8139]

I couldn't download the dataset. Can anyone share gdrive link of that dataset

[@sanjaypranav8481]

Use kaggle api command

[@radhikapatil8003]

Sir please do upload image classification using Fixefficientnet

[@radhikapatil8003]

@@jonsnow8297 what is this

[@pillivineeth5103]

Hi Krish!

[@mukhlisraza]

awesome tutorial

[@faresbecheikh7052]

Please how to plott the Confusion Matrix of this Project ?

[@jasimali3981]

i don't have a csv data for my sounds dataset , can i train use this model without csv please help me

[@bavithagurram8262]

Which model did you use?

[@jaaferklila1375]

hi how can i apply StandardScaler to the features

[@ayyoubm]

Thank you sir

[@muntazirmehdi503]

i have a question "why the first coefficient of MFCC has too much variance?" and what does each individual mfcc tells us? kindly reply soon as I ma not been able to find this answer, and my project is at halt due to this

[@aishamedina1505]

have you got the answer yet?

[@divyanshuawasthi6416]

file_name = os.path.join....... In this whole line i am getting key error 'fold'. Can you please help.

[@sanjaypranav8481]

U may not have included / this

[@niyatijain2119]

Did your error got solved?

[@dilipkn7339]

Same error I got

[@akanshasharma1358]

Same problem

[@satya-ni2rc]

Is there any data leakage issue In the code?

[@gayashandulanjana4025]

how can we create metadata csv file using audios

[@learningprogramming1308]

Can librosa support mp3?

[@gauranshluthra7520]

if 10 is the different classes we have, than what is 40???

[@dnyanwantnehare3250]

can anyone help me in In [75] how data is loading, Im working on anther sound data

[@louerleseigneur4532]

Thanks Krish

[@priyanshuprakash-k5w]

I am getting "No module named 'resampy'" error.Anyone knows how to fix this error.

[@032lovishkumar8]

hey , you were able to solve it?

[@ganapathisubramanyamjayam]

could you share the notebook?

[@anilkumarchannamsettti]

I have an error at file_ name

[@sylus121]

5:00 (Bookmark)

[@Gamezone-kq5sx]

Please tell me application of this project

[@AKrishnaAkhil]

dont repeat the content which was already explained in the previous vid, gets slightly confusing, even though it's not a major problem. like, during the start of the vid. the audio input, scipy and everything, already was explained in the first vid

[@sudhanwabokade1285]

11:16

[@hardishah3188]

why have you mentioned this timestamp, any thing to be noticed in particular?

[@sudhanwabokade1285]

@@hardishah3188 Just for my reference. It is supposed to be the time where I should resume watching the video.

[@anilkumarchannamsettti]

Pls help me

[@rupen1585]

Not useful to, very basic yoyu gave I think you have to make one more video how to calculate all the results, and all. for research point of view.