Detection-of-Plant-Disease
Detection-of-Plant-Disease copied to clipboard
kashish-ag
Increase the efficiency of the model
Calculate the current efficiency of the model will come approx 42% and try to increase its efficency.
I will like to work upon this issue. Can you please suggest which Algorithm should I focus on.
Hello, @vaibhavgupta2499 I have uploaded a new jupyter notebook ['PDD_Classification_Niranjan.ipynb'] in the main branch where I have incorporated a new approach by integrating the CNN with support vector machines & gradient boost model. And along with that, I have also included the further scope of improvement for the algorithm. So, Kindly check it out.
Sure, I will check it out and get back to you, as soon as possbile.
I am not able to work on this. Can you please reassign this issue to someone else who is more capable than me.
@kashish-ag ya it is assigned to someone but I think I can resolve this issue.@kashish-ag if by chance @Sumit11081996 can't resolve this issue may you then assign this issue to me.
Sure Dhruv I'll assign it to you if @Sumit11081996 can't. Also, @Sumit11081996 what's the progress?
I am working on it if the accuracy is not increase then @dhruv-varshney will go for it.
@kashish-ag where you calculated the efficiency since i can not able to find that 42%
@dhruv-varshney I have built a hybrid model using CNN and XGB in the 'PDD_Classification_Niranjan.ipynb' Jupyter Notebook. There I have also mentioned the further scope. Try checking it out.
@dhruv-varshney Since there are many peaks in the graph as well try playing with the optimizer and other hyperparameters. If you get an OSError saying that you have run out of memory. Just scale down the batch size and you will be good to go.
But so far when I personally tested on my system it was working fine but when I was trying to run it on colab instance it was crashing. So try to create a data pipeline so that you push the images in batches to the GPU, in that way you won't be crashing the instance.
Also, if possible try to take a video of any plant near your vicinity/online, then scale down the model to tf-lite and do a real time analysis.
I am doing with trial and error and I am finding the values in which we get the result in less time. But most of the time I am getting the same efficiency. still, I am trying, but I need time.