search

pankhurivanjani / Extended_plant_disease_detecttion

2 stars 0 forks source link

Select appropriate dataset for experimentation (Select more recent dataset with proper testset and benchmark) #3

Closed nik1806 closed 4 years ago

nik1806 commented 4 years ago

Also check license

pankhurivanjani commented 4 years ago

https://github.com/pankhurivanjani/Extended_plant_disease_detecttion/blob/WIP-pankhuri/Dataset/Selection.md

nik1806 commented 4 years ago

Based on reviews of datasets, here are my comments for further steps:

  1. Usually, a larger dataset is required for training with Deep learning techniques, so preference should be given to larger size.
  2. Yes, it is possible to combine two datasets. It is common to use an additional similar dataset while training to improve the results on test set.
  3. For now, let's focus on single leaf case. If time permits we can solve multiple leaves case.

In summary, choose the dataset having crops/veggies/fruits most similar to plant village dataset considering the above recommendations. (We will use augmented dataset as reference for comparing our own augmentation, similarly with Kaggle notebooks.)

nik1806 commented 4 years ago

@pankhurivanjani Please read this paper also to decide. https://github.com/pankhurivanjani/Extended_plant_disease_detecttion/issues/2#issuecomment-668392629