Indian Truck Detection Overfitting

[aryan1113]

According to the project readme, the accuracy of model is 100%, which seems to good to be true.

The following changes are proposed

1. Implementing dropout layers, so that neurons learn independently
2. Using a simpler architecture, instead of the 512 neuron dense layer.
3. Making use of a better dataset, original dataset comprises only of ~500 images, all of which correspond to an Indian Truck, none of them are of any other label.

[abhisheks008]

You want to enhance this whole project of Indian Truck Detection right?
@aryan1113

[yash16bhargava]

I want to work on this issue.

[abhisheks008]

@yash16bhargava what do you want to do?

[abhisheks008]

@aryan1113 @yash16bhargava any update on this issue?

[aryan1113]

Sorry for the confusion, I saw a very obvious overfitting issue with the model made by @PiyushBL45t and suggested improvements for the same.
I didn't intend to work on this issue.

@yash16bhargava Maybe something along the lines of data augmentation should help.
Augmentation is making use of the properties of images to create new images, namely through rotating, scaling and shifting images.
Also try out some of the methods that are mentioned In the original issue : )

[veer-chheda]

I'd like to work on this issue by using a combination of data augmentation, reducing the network complexity and dropout layers.
@abhisheks008

[abhisheks008]

One issue at a time @veer-chheda

[Akansha-Mulchandani]

I'd like to work on the data augmentation and the dropout part!

[abhisheks008]

Hi @Akansha-Mulchandani this is an existing project, https://github.com/abhisheks008/DL-Simplified/tree/main/Indian%20Truck%20Detection

As you can go through the project you can find that the project's models are overfitted as a result they are showing 100% accuracy. Your job is to add TWO MORE deep learning models for this same dataset and fine tune the existing two models.

No need to create a separate folder for your work. You have to push your files in the existing project folder of Indian Truck Detection.

Are you up for this?

[Akansha-Mulchandani]

yep i'll take this up

[abhisheks008]

Assigned @Akansha-Mulchandani

[Akansha-Mulchandani]

Hey Abhishek! I looked at the dataset and it’s not labelled, do you have one with the labels or do you need me to do it ?

…

On Sat, 25 May 2024 at 9:53 PM, Abhishek Sharma ***@***.***> wrote: Assigned #275 <#275> to @Akansha-Mulchandani <https://github.com/Akansha-Mulchandani>. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA> . You are receiving this because you were assigned.Message ID: ***@***.***>

[abhisheks008]

Hey Abhishek! I looked at the dataset and it’s not labelled, do you have one with the labels or do you need me to do it ?
…
On Sat, 25 May 2024 at 9:53 PM, Abhishek Sharma @.> wrote: Assigned #275 <#275> to @Akansha-Mulchandani https://github.com/Akansha-Mulchandani. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA . You are receiving this because you were assigned.Message ID: @.>

But the same dataset is used for the existing project right? If the existing dataset is working with your model implementation then you can take another dataset for the enhancement model implementation.

[Akansha-Mulchandani]

Hey Abhishek, I'm done with the finetuning and two models,let me know if you would like to have a look at the code or should I add it directly ? On Tue, 28 May 2024 at 09:15, Abhishek Sharma ***@***.***> wrote:

…

Hey Abhishek! I looked at the dataset and it’s not labelled, do you have one with the labels or do you need me to do it ? … <#m_-4476837291662388748_> On Sat, 25 May 2024 at 9:53 PM, Abhishek Sharma *@*.*> wrote: Assigned #275 <#275> <#275 <#275>> to @Akansha-Mulchandani <https://github.com/Akansha-Mulchandani> https://github.com/Akansha-Mulchandani <https://github.com/Akansha-Mulchandani>. — Reply to this email directly, view it on GitHub <#275 (comment) <#275 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA <https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA> . You are receiving this because you were assigned.Message ID: @.*> But the same dataset is used for the existing project right? If the existing dataset is working with your model implementation then you can take another dataset for the enhancement model implementation. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A5RWN3HS7NZXKRFJHBSZJITZEP4XHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZUGI4TKNZRGA> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[abhisheks008]

Hey Abhishek, I'm done with the finetuning and two models,let me know if you would like to have a look at the code or should I add it directly ? On Tue, 28 May 2024 at 09:15, Abhishek Sharma @.> wrote:
…
Hey Abhishek! I looked at the dataset and it’s not labelled, do you have one with the labels or do you need me to do it ? … <#m_-4476837291662388748_> On Sat, 25 May 2024 at 9:53 PM, Abhishek Sharma @.> wrote: Assigned #275 <#275> <#275 <#275>> to @Akansha-Mulchandani https://github.com/Akansha-Mulchandani https://github.com/Akansha-Mulchandani https://github.com/Akansha-Mulchandani. — Reply to this email directly, view it on GitHub <#275 (comment) <#275 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA . You are receiving this because you were assigned.Message ID: @.> But the same dataset is used for the existing project right? If the existing dataset is working with your model implementation then you can take another dataset for the enhancement model implementation. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3HS7NZXKRFJHBSZJITZEP4XHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZUGI4TKNZRGA . You are receiving this because you were mentioned.Message ID: @.**>

Create your pull request, I'll review it there.

[Akansha-Mulchandani]

pull request sent #775 On Sun, 2 Jun 2024 at 11:48, Abhishek Sharma ***@***.***> wrote:

…

Hey Abhishek, I'm done with the finetuning and two models,let me know if you would like to have a look at the code or should I add it directly ? On Tue, 28 May 2024 at 09:15, Abhishek Sharma *@*. *> wrote: … <#m_-5853466010497699131_> Hey Abhishek! I looked at the dataset and it’s not labelled, do you have one with the labels or do you need me to do it ? … <#m_-4476837291662388748_> On Sat, 25 May 2024 at 9:53 PM, Abhishek Sharma @.> wrote: Assigned #275 <#275> <#275 <#275>> <#275 <#275> <#275 <#275>>> to @Akansha-Mulchandani <https://github.com/Akansha-Mulchandani> https://github.com/Akansha-Mulchandani <https://github.com/Akansha-Mulchandani> https://github.com/Akansha-Mulchandani <https://github.com/Akansha-Mulchandani> https://github.com/Akansha-Mulchandani <https://github.com/Akansha-Mulchandani>. — Reply to this email directly, view it on GitHub <#275 <#275> (comment) <#275 (comment) <#275 (comment)>>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA <https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA> https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA <https://github.com/notifications/unsubscribe-auth/A5RWN3BFEV4MP3NM2T5JVGLZEC3IHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV45UABCJFZXG5LFIV3GK3TUJZXXI2LGNFRWC5DJN5XDWMJSHEZTGNBUGY3TAOA> . You are receiving this because you were assigned.Message ID: @.> But the same dataset is used for the existing project right? If the existing dataset is working with your model implementation then you can take another dataset for the enhancement model implementation. — Reply to this email directly, view it on GitHub <#275 (comment) <#275 (comment)>>, or unsubscribe https://github.com/notifications/unsubscribe-auth/A5RWN3HS7NZXKRFJHBSZJITZEP4XHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZUGI4TKNZRGA <https://github.com/notifications/unsubscribe-auth/A5RWN3HS7NZXKRFJHBSZJITZEP4XHAVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMZUGI4TKNZRGA> . You are receiving this because you were mentioned.Message ID: @.***> Create your pull request, I'll review it there. — Reply to this email directly, view it on GitHub <#275 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/A5RWN3E7FPNK6A4ILMLNHULZFK2L3AVCNFSM6AAAAAAY6FZH3KVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCNBTG4YTMMZWGU> . You are receiving this because you were mentioned.Message ID: ***@***.***>

[github-actions]

Hello @Akansha-Mulchandani! Your issue #275 has been closed. Thank you for your contribution!