The problem KRISHI-TECH solves

The KRISHI-TECH platform as one stop destination for farmers with the help of machine learning solves the mojor problems faced by farmers like what fertilizer should use for his yields, growing crop that maximize the profit of farmers and predicting disease in his field , the two major problems solved by KRISHI-TECH platform are as follows

1. Educate the farmers : The disease prediction model can predict the disease that are not local to particular area in response this will help the farmer to correctly choose the fertilizer or medicine according to the disease type.
   The blog section on the KRISHI-TECH further educated the farmer about the latest happening in the world of agriculture

2. CHAT-BOT : The intigrated chatbot on the KRISHI-TECH platform can solve the queries of the farmer using the diverse knowledge base and artificial intelligence, furthermore it can suggest government schemes and financial assistance to farmers resulting in reduced stress among the farmer community and maximizing the profits

3. MARKET PLACE: The integrated market place acts as the intermediate between buyers from different regions and farmers resulting in increasing demand for farmer and as this platform will be connecting to some unknown buyers to the farmers

Challenges we ran into

1. CHATBOT: Training the chatbot was one of the difficult jobs , as it need to be trained on our custom data and more over a research work was required and information and resource material from government database, as to solve the complex queries from farmers that may include latest goverment schemes providing financial assistance , loans from nearest banks that have very low interest rate that suites the farmer.
   2.DATASET: We have majorly used three datasets for crop recommendation and fertilizer recommendation and disease prediction. Firstly finding the accurate dataset and correcting the incorrect values was a huge task as dataset was very complex and large.

3.BACK-END AND FRONT-END CONNECTIVITY: Handling flask framework of python was a tough job as there was problem of hashing authorization as a result we were not getting the desired output and hence the accuracy of the model was not up to the mark, with the help of online resources , blogs we solved the issue and the accuracy of the model was updatad to 85%.

4.CNN: As CNN was new to us we faced many issues with the training of the model using CNN,
a. Low accuracy
b. computational issue and training time
c. archietecture freez and unfreez