Potato Disease Classification Project | Convolutional Neural Network (CNN)

Potato Disease Classification Project | Convolutional Neural Network (CNN) | Deep Learning

Рет қаралды 484

Ай бұрын

Hey everyone,
In this video I have implemented and explained a Convolutional Neural Network (CNN) project using python, tensor flow and keras.
Dataset : www.kaggle.com/datasets/faysa...
IDE : Google Colab - colab.google
#deeplearning
#cnn
#python
#tensorflow
#keras
Thanks for watching!
Please LIKE, SHARE and SUBSCRIBE if you enjoyed the content.

Пікірлер: 9

@RDFunFactory563 12 күн бұрын

outstanding work, I learned so many things from your DL videos, can you please make a video on augmentation data with transfer learning using the image dataset you used in the lung cancer detection. thank you!

@ashpritmehra8174 Ай бұрын

hi thanks for this, what is the best way to learn DL quickly for classification like this?

@InsightsByRish Ай бұрын

Welcome 😊. If your theoretical concepts are clear then I would say the best way is to watch a lot of implementation videos where every person explains with different methods. At the end of the day you at least get the core gist of things altogether.

@pratikraut1376 25 күн бұрын

Thank you for the video, I trained it for more epochs and the result is --> in terms of accuracy it improves and after a few epochs the accuracy decreases and again it improves, it is imbalance but for only 1 class (healthy) so, does this kind of behaviour of model shows overfitting or some other problem ?

@InsightsByRish 25 күн бұрын

I think it's likely a combination of both factors(overfitting and imbalance). But still I would say 'Overfitting' can highly be a factor (as that's a small dataset getting trained on high no.of epochs).So you might try adding few dropout layers or Batch normalisation layer. After adding that your accuracy will be dropped a bit but atleast it wouldn't fluctuate.

@pratikraut1376 22 күн бұрын

@@InsightsByRish After experimenting with parameters, it is not fluctuating --> I reduced the number of data augmentation techniques, changed the learning rate to 0.002 and also changed strides to 1. It is still overfitting though may be the dataset is small need to work on it.