Scribd
Upload
89 views10 pages

Deep Audio Classification

This document provides an overview of a mini project to classify bird calls using convolutional neural networks and TensorFlow. The goal is to build a model that can count the number of Capuchin bird calls within an audio clip. The team will take audio clips containing both Capuchin and non-Capuchin bird calls, convert them to spectrograms, and use a CNN model to classify each 3-second clip as containing a Capuchin call or not. The trained model will then be used to analyze longer forest recordings split into 3-second windows to identify and group consecutive Capuchin calls to calculate the density of Capuchin birds within the recording.

Uploaded by

Vinay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
89 views10 pages

Deep Audio Classification

This document provides an overview of a mini project to classify bird calls using convolutional neural networks and TensorFlow. The goal is to build a model that can count the number of Capuchin bird calls within an audio clip. The team will take audio clips containing both Capuchin and non-Capuchin bird calls, convert them to spectrograms, and use a CNN model to classify each 3-second clip as containing a Capuchin call or not. The trained model will then be used to analyze longer forest recordings split into 3-second windows to identify and group consecutive Capuchin calls to calculate the density of Capuchin birds within the recording.

Uploaded by

Vinay
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 10

MINI PROJECT

SESSION NO: 1​
DATE: 27/08/2022​
REVIEW NO: ZERO​
BATCH: 15
DEEP AUDIO
CLASSIFICATION
Using Tensorflow and Convolutional
neural networks
TEAM MEMBERS

• SHREE VARSHAN B
• YASHWANT B
• PUNEETH RAM P
• VINAY NARAYAN
PROBLEM STATEMENT

• This is one of the challenges proposed by HP unlocked and it's about deep
audio analysis.​

• The Challenge is to build a Machine Learning model and code to count the
number of Capuchin bird calls within a given clip.​
INTRODUCTION

• The dataset comprises of bird calls in the form of audio clips out of which
the density of Capuchin bird calls has to be measured using Convolutional
neural networks and tensorflow.​

• Python is used to build the tensorflow CNN model and to identify whether
the clip contains Capuchin bird call.​
SOLUTION

• The training data comprises of audio clips of bird calls of both capuchin
bird calls and non-capuchin bird calls which is passed through the Cnn
model after being converted to spectrogram form from the provided audio
waveforms.​

• We can then apply computer vision techniques on the spectrograms to


perform the classification​
• The model output hands out a binary data [0,1] where 1 represents that a
capuchin bird is heard in the clip and 0 for no capuchin bird call in the
particular clip.​

• The forest recordings of roughly 3 minutes each is split into 3 second


segment windows.​

• Then the trained Tensorflow neural network is slid across each of these
windows to find the capuchin bird calls.​

• Finally the consecutive capuchin bird calls are grouped together as one as
the belong to the same bird and the density of capuchin bird call in the
particular recording is calculated.​​
ARCHITECTURE DIAGRAM

Bird call clips Read as Waveforms CNN model


(3 seconds) waveform using converted into using
tensorflow spectrograms tensorflow

Binary
output

1 if capuchin bird call is detected


within the particular clip,
0 otherwise
3 second​ 3 second​ 3 second​ 3 second​ 3 second​ 3 second​

3 min forest recording split into 3


Trained Tensorflow neural network
second windows
model

Output
0 1 1 0 1 1

consecutive calls consecutive calls


grouped into a single grouped into a single
capuchin call capuchin call
THANK YOU

You might also like