Final project for the Building AI course
Detection of pests and diseases in plants using AI and suggesting organic pesticide and management techniques. Kira (कीरा) means insect in Nepali.
Farmers have to deal with following problems in regards to pests and diseases:
- Cannot easily identify new pests or diseases without the help of experts
- Don't know about effective management techniques to deal with pests and diseases
- Overuse of chemical insecticides and pesticides, which leads adverse effect on the consumers and increases resistance in the pests
- Lack of knowledge about Integrated Pest Management techniques
- Lack of warning on potential outbreak of pests and diseases in their areas based upon weather patterns and new outbreaks
These problems are very prevalent in Nepal and my guess is that the situation is similar around the world. With this app I hope to help farmers in rural areas in identifying diseases and pests and help manage them effectively with novel non-destructive techniques and reap the most benefit from their crops.
Describe the process of using the solution. In what kind situations is the solution needed (environment, time, etc.)? Who are the users, what kinds of needs should be taken into account? Farmers will use their smartphone's camera to take the photo of the pest or disease in their crop. Then based upon the trained model, the app will identify the pest/disease and provide info about it and suggest techniques to manage it. Based upon farmers data about the location and crop cycle planning, the app will use data sourced from other platforms for weather and plant disease outbreaks and make predictions if it will affect the area and warn the farmer accordingly.
Users will be farmers in mostly rural areas. Farmers mostly are not tech savyy so the UI and UX needs to be as simple and intuitive as possible. For the product to be viable, apart from the AI component, equal effort must be invested in doing UX research preferably through Human Centered Design approach such as Design Thinking.
For initial training data can be gathered from existing sources such as iNaturalist app and other open data sources. More work is needed to identify sources of good data. After the app is launched, the data submitted by the users in form of photos, location info and cro info etc will be used to continuously train and refine the models.
Need more brainstorming on this topic. TODO!
Right now, it is just an idea. More research needs to be done on data sources, potential users, use conditions, technical roadmap and content. Any contribution in any form is always welcome.