using UnityEngine;
using System.Collections;
using UnityEngine.XR;
using System.IO;
using System;
using System.Linq;
using System.Collections.Generic;
using System.Text;
using UnityEngine.Networking;
using UnityEngine.UI;
using System.Threading;

using TensorFlow;


public class AccelerometerInputCNN : MonoBehaviour
{
	// set by trained CNN model
	public int inputWidth = 90;

	// third value corresponds to inputWidth
	public TextAsset graphModel;
  // the dimenstions of 'inputTensor' correspond to the dimensions of the input for my trained graph
  private float[,,,] inputTensor = new float[1, 1, 90, 1];

	// determine if person is walking from cnn returned value
	private int standIndex = 0;

	// how many options of activities we have - standing, walking, jogging
	private int activityIndexChoices = 2;

  // list to collect acceleration values
  private List<float> accelL;

	float confidence = 0;
	float sum = 0f;
	float test = 0f;
	int activity = 0;
  int index = 0;

	void Start ()
	{
		#if UNITY_ANDROID
		TensorFlowSharp.Android.NativeBinding.Init ();
		#endif

    // initialize the oculus go display
		display = new OVRDisplay ();

		accelL = new List<float> ();
	}

	void FixedUpdate ()
	{
    collectValues();
	}

  void collectValues()
  {
        float curr = display.acceleration.y;

        if (accelL.Count < inputWidth) {
            accelL.Add (curr);
        }
        if (accelL.Count == inputWidth) {
            index = evaluate ();
            accelL.RemoveAt (0);
            accelL.Add (curr);
        }
  }

  void evaluate ()
  {
      // convert from list to array
      for (int i = 0; i < accelL.Count; i++) {
          inputTensor [0, 0, i, 0] = accelL [i];
          test = inputTensor [0, 0, i, 0];
      }

      // declare a list for outputs
      float[,] recurrentTensor;

      // set up and run a graph session
      using (var graph = new TFGraph())
      {
          graph.Import(graphModel.bytes);
          var session = new TFSession(graph);

          var runner = session.GetRunner();

          // implicitally convert a C# array to a tensor
          TFTensor input = inputTensor;

          // set up input tensor and input
          // KEY: I am telling my session to go find the placeholder named "input_placeholder_x" and user my input TENSOR instead
          runner.AddInput(graph["input_placeholder_x"][0], input);

          // set up output tensor
          runner.Fetch(graph["output_node"][0]);

          // run model - recurrentTensor now holds the probabilities for each outcome
          recurrentTensor = runner.Run()[0].GetValue() as float[,];

          // frees up resources - very important if you are running graph > 400 or so times
          session.Dispose();
          graph.Dispose();
      }

      // get the option with the highest probability
      float highVal = 0;
      int highInd = -1;
      sum = 0f;

      for (int j = 0; j < activityIndexChoices; j++) {
          confidence = recurrentTensor [0, j];
          if (highInd > -1) {
            if (recurrentTensor [0, j] > highVal) {
              highVal = confidence;
              highInd = j;
            }
          } else {
            highVal = confidence;
            highInd = j;
          }

          // debugging - sum should = 1 at the end
          sum += confidence;
       }

      // save the index of the highest value for later use!
     index = highInd;
  }