The problem AI Based Crop Identifier solves

Currently there are not many AI based crop identifier. This can lead to wrong predictions as there are chances of manual or machine errors. The model can help identifying crops like potato and ginger which are difficult to identify because they grow underground. An extension of this can be that the model also shows the type of soil, weather condition and fertilizzers that can help in its growth the most. Not only food and cash crops, It can be trained to identify flowers and this can help beekeepers to let their bees get best flowers and make honey.
Agriculture being a big domain, It is necessary to focus on its advancement. It is linked to each and every person.
The model can help farmers, Researchers, Gardeners, Beekeepers and even other people and nature enthusiasts.

Challenges I ran into

The biggest challenge in any ML/DL model is collection of Data. I had to get 300 pictures of about 25 crops and that too the actual pictures with no farmers, humans, cars, or things like that. I used web scrapers to overcome this. Once i got the pictures, I had to manually delete the unfit pictures.
Second challenge was to avoid overfitting and underfitting. I had to train my model extensively.
Also I had to try the model wil many architectures like resnet 156, resnet 50 and googlenet and etc to find the most apt one.