An app with deep learning model to predict compatibility using soil images and also assists in minimising loss at harvest time by predicting prices. Developed at IndiaHacks 2017
- Beautiful Soup
- Tensorflow
- Numpy
- Pandas
- First scrape crop prices data from agmarknet.gov.in using python web scraping website beautiful soup.
python scrap_crops.py
This will give output csv file for different states, crops and over different months.
- To predict prices of crops after harvesting period, run
python ./scripts/hack1.py
This uses a random forest regressor to predict prices.
- Train an inceptionv3 model inn tensorflow using transfer learning to classify soil type by image.
python ./scripts/retrain.py --imagedir <pathtoimages>
- Complete pipeline output\ Get best crop to grow from soil image taking into consideration price prediction after harvest time.
python ./scripts/label_image.py