!wget https://github.com/fazalmittu/PatchDetection/raw/master/BlackedOutLawn.jpg
!wget https://github.com/fazalmittu/PatchDetection/raw/master/BlackedOutLawn_Detected.jpg
!wget https://github.com/fazalmittu/PatchDetection/releases/download/v3.0/detection_model-ex-04--loss-25.86.h5
!wget https://github.com/fazalmittu/PatchDetection/releases/download/v3.0/detection_config1700_v1.json
!wget https://github.com/fazalmittu/PatchDetection/raw/master/BlackedOutFullLawn.jpg
!wget https://github.com/fazalmittu/PatchDetection/raw/master/BlackedOutFullLawn_Detected.jpg
!wget https://github.com/fazalmittu/PatchDetection/raw/master/SprinklerPOV.jpg
!wget https://github.com/fazalmittu/PatchDetection/raw/master/SprinklerPOV_Detected.jpg
!wget https://github.com/OlafenwaMoses/ImageAI/releases/download/essential-v4/pretrained-yolov3.h5

!pip uninstall -y tensorflow
!pip install tensorflow-gpu==1.13.1
!pip install keras==2.2.4
!pip install imageai==2.1.0
!pip install numpy

%load_ext autoreload
%autoreload 2
from google.colab import drive
import sys
from pathlib import Path
drive.mount("/content/drive", force_remount=True)
base = Path('/content/drive/MyDrive/PatchDetectorProject/')
sys.path.append(str(base))
zip_path = base/'AugmentedDataSetFinal.zip'
!cp "{zip_path}" .
!unzip -q AugmentedDataSetFinal.zip
!rm AugmentedDataSetFinal.zip

from imageai.Detection.Custom import DetectionModelTrainer
from __future__ import print_function
trainer = DetectionModelTrainer()
trainer.setModelTypeAsYOLOv3()
trainer.setDataDirectory(data_directory="AugmentedDataSetFinal")
trainer.setTrainConfig(object_names_array=["patch"], batch_size=4, num_experiments=5, train_from_pretrained_model="pretrained-yolov3.h5")
trainer.trainModel()

from imageai.Detection.Custom import DetectionModelTrainer
trainer = DetectionModelTrainer()
trainer.setModelTypeAsYOLOv3()
trainer.setDataDirectory(data_directory="AugmentedDataSetFinal")
trainer.evaluateModel(model_path="AugmentedDataSetFinal/models", json_path="AugmentedDataSetFinal/json/detection_config.json", iou_threshold=0.5, object_threshold=0.3, nms_threshold=0.5)

from imageai.Detection.Custom import CustomObjectDetection
from PIL import Image, ImageDraw
import numpy as np
detector = CustomObjectDetection()
detector.setModelTypeAsYOLOv3()
detector.setModelPath("/content/detection_model-ex-04--loss-25.86.h5")
detector.setJsonPath("/content/detection_config1700_v1.json")
detector.loadModel()
detections = detector.detectObjectsFromImage(input_image="SprinklerPOV.jpg", output_image_path="SprinklerPOV_Detected.jpg", minimum_percentage_probability=30)
i = 0
coord_array = []
for detection in detections:
    coord_array.append(detection["box_points"])
    print(detection["name"], " : ", detection["percentage_probability"], " : ", detection["box_points"])
    i+=1
print(coord_array)
detected = Image.open("SprinklerPOV_Detected.jpg")
box = ImageDraw.Draw(detected)
for i in range(len(coord_array)):
  box.rectangle(coord_array[i], width=10)
detected

!wget https://github.com/fazalmittu/PatchDetection/raw/master/FeetToPixel.JPG
img_ft = Image.open("FeetToPixel.JPG")
ft_line = ImageDraw.Draw(img_ft)
ft_line.line([(175, 1351), (362, 1360)], fill=(0, 255, 0), width=10)
ft_distance = np.sqrt(9*9 + 187*187)
print(ft_distance)
img_ft

from PIL import Image, ImageDraw
#TOP LEFT = [0, 1]
#BOTTOM LEFT = [0, 3]
#TOP RIGHT = [2, 1]
#BOTTOM RIGHT = [2, 3]
img = Image.open("SprinklerPOV_Detected.jpg")
middle_line = ImageDraw.Draw(img)
avg_1Line = ImageDraw.Draw(img)
avg_2Line = ImageDraw.Draw(img)
avg_1 = (coord_array[1][1] + coord_array[1][3])/2
avg_2 = (coord_array[1][0] + coord_array[1][2])/2
middle_line.line([(2180, 0), (2180, 3024)], fill=(0, 255, 0), width=10)
# avg_1Line.line([(coord_array[1][0], coord_array[1][1]), (coord_array[1][0], coord_array[1][3])], fill=(255, 0, 0), width=10)
# avg_2Line.line([(coord_array[1][0], coord_array[1][3]), (coord_array[1][2], coord_array[1][3])], fill=(255, 0, 0), width=10)
def find_angle():
  line_to_patch = ImageDraw.Draw(img)
  line_to_patch.line([(avg_2, avg_1), (2180, 3024)], fill=(255, 0, 0), width=10)
  length_1_vertical = 3024 - avg_1
  length_2_horizontal = 2500 - avg_2
  print("Distance = ", np.sqrt(length_1_vertical*length_1_vertical + length_2_horizontal*length_2_horizontal)/ft_distance, "ft")
  angle_radians = np.arctan(length_2_horizontal/length_1_vertical)
  angle = (180/(np.pi/angle_radians)) #Convert radians to degrees
  return angle
print(avg_1, avg_2)
print(find_angle())
img

/*
*
* Fazal Mittu; Sprinkler Control
*
*/
const int ROTATEstepPin = 3;
const int ROTATEdirPin = 4;
const int ANGLEstepPin = 6;
const int ANGLEdirPin = 7;
const int ROTATEangle = 42.25191181;//TODO: Find Conversion: Steps --> Angle 1000 steps = 90 degrees
const int ANGLEangle = 12.76187539;// pixels --> ft: 187 pixels = 1 feet
bool TURN = true;
float angleToSteps(float angle){
  float steps = 1000/(90/angle);
  return steps;
}
float ftToSteps(float feet) {
  float steps = 100/(8/feet);
  return steps;
}
void setup() {
  Serial.begin(9600);
  pinMode(ROTATEstepPin,OUTPUT);
  pinMode(ROTATEdirPin,OUTPUT);
  pinMode(ANGLEstepPin,OUTPUT);
  pinMode(ANGLEdirPin,OUTPUT);
}
void loop() {
  int ROTATEsteps = angleToSteps(ROTATEangle); //Angle was determined using Python Script
  int ANGLEsteps = angleToSteps(ANGLEangle);
  delay(7000);
  if (TURN == true) {
    for(int x = 0; x < ROTATEsteps; x++) {
      digitalWrite(ROTATEstepPin,HIGH);
      delayMicroseconds(500);
      digitalWrite(ROTATEstepPin,LOW);
      delayMicroseconds(500);
    }
    delay(5000);
    for (int x = 0; x < 100; x++) { //100 steps = 8 ft, 0 steps = 14.5 ft
      digitalWrite(ANGLEstepPin,HIGH);
      delayMicroseconds(500);
      digitalWrite(ANGLEstepPin,LOW);
      delayMicroseconds(500);
    }
  }
  TURN = false;
}