NN hosted as service using Flask and JS
This simple classifier uses transfer learning to train an existing VGG16 model and by changing the last layer we create a new model which predicts if the uploaded image is cat or dog
Note : The actual model is more than 500 MB so , have not hosted it on Github
After placing the model in repective directory start the Flask server using
- export FLASK_APP=predict_app.py
- flask run --host=0.0.0.0
Change the IP address to point to your webserver
To get predicitons from shell instead of browser
$fileName='cat.png'
$bytes = [IO.File]::ReadAllBytes($fileName)
$base64Image = [Convert]::ToBase64String($bytes)
$message = @{ image = $base64Image }
$jsonified = ConvertTo-Json $message
$response = Invoke-RestMethod -Method Post -Url "http://0.0.0.0:5000/predict" -Body $jsonified
$response.prediction | format-listCredits : This was build using tutorials from this DeepLizard Course