Feature : Support .tflite Predictions

[kshitijrajsharma]

• Supprots tflite predictions which will remove the dependency from huge tensorflow making it very light and easy for deployment of inference module

**How to test ? **

• Download .tflite model
• Change extension to tests/app_test.py inlcude your .tflite , only install tflite-runtime , Don't install whole tensorflow , Try to run the prediction

Example :

• Install fairpredictor
• Install tflite-runtime
• Download .tflite model from fAIr

Run Following

**from predictor import predict

model_path = "checkpoint.tflite"
bbox = [100.56228021333352, 13.685230854641182, 100.56383321235313, 13.685961853747969]
zoom_level = 20
tms_url = "https://tiles.openaerialmap.org/6501a65c0906de000167e64d/0/6501a65c0906de000167e64e/{z}/{x}/{y}"

bbox = [100.56228021333352, 13.685230854641182, 100.56383321235313, 13.685961853747969]
zoom_level = 20
tms_url = "https://tiles.openaerialmap.org/6501a65c0906de000167e64d/0/6501a65c0906de000167e64e/{z}/{x}/{y}"

my_predictions = predict(bbox, model_path, zoom_level, tms_url, remove_metadata=False)
print(my_predictions)
**

TODO

• Clean up the predict.py , currently its very messy and lot of repetition of code