The inspiration of deeplant.ai is to combine agriculture and technology using artificial intelligence. Where artificial intelligence in the form of machine learning is made with the help of python and tensorflow.
What it does
deeplant.ai serves to provide information on plants or plantations in real time, and is able to detect pests or diseases in plants making it easier to overcome. This will help increase agricultural production in the future.
How I built it
we started by making a model using python and backend tensorflow with google colabs to training the data. Our data is taken secondary from several open sources, especially Kaggle. Then we save the model obtained in the form of .tflite file and then it is implemented in the Android application to detect diseases and do other things about agriculture
Challenges I ran into
the challenges that we have encountered, especially on computers that lack support to develop applications and impact on slowing our performance and we found some error with code or application. then the data is a little difficult to find. Knowledge in several programming languages at once is still lacking. Changing from .flite to tensorflow.js is a bit complicated and also the application of tensorflow with the Internet of Things (IoT) has not been implemented due to the limitations of the above.
Accomplishments that I'm proud of
Until now, we have successfully created a model and saved it as a .tflite file and implemented it in the application and have been able to apply it to tensorflow.js. besides that we try to make a smooth framework for deeplant.ai applications
What I learned
In its application we learned a lot that tensorflow 2.0 is easier to use than version 1, and produces better output even though we have to adjust the changes and learn more, but we are happy with that. and we think tensorflow will be widely used to help many people in the future, especially the application of AI
What's next for deeplant.ai
The next step, we will apply to tensorflow.js so that it is more flexible than the .tflite file, then we also intend to develop this application for the IoT. In addition to getting data directly and funding to overcome problems, we want to try to work with various parties including researchers, the government and its expectations with Google and Tensorflow