VARUN SHARMAN 1660484Jival Jenson
Published © GPL3+

Ichigo: The Agricultural Bot

Ichigo is an agricultural robot that goes around the cultvational land and gets you statistics about the diseased plants and the field.

AdvancedFull instructions provided10 hours1,292
Ichigo: The Agricultural Bot

Things used in this project

Story

Read more

Schematics

Connections

Code

Processing code

C/C++
This is the processing code to send the values to the bluetooth module of arduino
import processing.serial.*;

Serial myPort;  // Create object from Serial class

void setup() 
{
  size(200,200); //make our canvas 200 x 200 pixels big
  println(Serial.list());
  String portName = Serial.list()[0]; //change the 0 to a 1 or 2 etc. to match your port
  myPort = new Serial(this, portName, 9600);
}
void draw() 
{
  
  int count=0;
  if (keyPressed == true) 
  {   
    if (key=='d')
  { myPort.write("d");         
   println(key);  
   delay(100);
  }
  if (key=='s')
  { myPort.write("s");         
   println(key);  
   delay(100);
  }
  if (key=='a')
  { myPort.write("a");         
   println(key);  
   delay(100);
  }
  if (key=='w')
  { myPort.write("w");         
   println(key);  
   delay(100);
  }
  if (key=='i')
  { myPort.write("i");         
   println(key);  
   delay(100);
  }
  if (key=='k')
  { myPort.write("k");         
   println(key);  
   delay(100);
  }
  if (key=='j')
  { myPort.write("j");         
   println(key);  
   delay(100);
  }
  if (key=='l')
  { myPort.write("l");         
   println(key);  
   delay(100);
  }
  if (key=='c')
  { myPort.write("c");         
   println(key);  
   delay(100);
  }
  } 
  else 
  println("__INVALID/NO INPUT__");
}

Sony Board Code - Camera, GPS, and Ultrasonic Sensor

C/C++
This is the code that runs on the robot and includes code for the camera, bluetooth, ultrasonic sensor and the GPS
#include <SDHCI.h>
#include <stdio.h>  /* for sprintf */

#include <Camera.h>
#include <GNSS.h>

#define STRING_BUFFER_SIZE  128       /**< %Buffer size */

#define RESTART_CYCLE       (60 * 5)  /**< positioning test term */

static SpGnss Gnss;                   /**< SpGnss object */

#define BAUDRATE                (9600)

/**
 * @enum ParamSat
 * @brief Satellite system
 */
enum ParamSat {
  eSatGps,            /**< GPS                     World wide coverage  */
  eSatGlonass,        /**< GLONASS                 World wide coverage  */
  eSatGpsSbas,        /**< GPS+SBAS                North America        */
  eSatGpsGlonass,     /**< GPS+Glonass             World wide coverage  */
  eSatGpsQz1c,        /**< GPS+QZSS_L1CA           East Asia & Oceania  */
  eSatGpsGlonassQz1c, /**< GPS+Glonass+QZSS_L1CA   East Asia & Oceania  */
  eSatGpsQz1cQz1S,    /**< GPS+QZSS_L1CA+QZSS_L1S  Japan                */
};

/* Set this parameter depending on your current region. */
static enum ParamSat satType =  eSatGps;

/**
 * @brief Turn on / off the LED0 for CPU active notification.
 */
static void Led_isActive(void)
{
  static int state = 1;
  if (state == 1)
  {
    ledOn(PIN_LED0);
    state = 0;
  }
  else
  {
    ledOff(PIN_LED0);
    state = 1;
  }
}

/**
 * @brief Turn on / off the LED1 for positioning state notification.
 *
 * @param [in] state Positioning state
 */
static void Led_isPosfix(bool state)
{
  if (state)
  {
    ledOn(PIN_LED1);
  }
  else
  {
    ledOff(PIN_LED1);
  }
}

/**
 * @brief Turn on / off the LED3 for error notification.
 *
 * @param [in] state Error state
 */
static void Led_isError(bool state)
{
  if (state)
  {
    ledOn(PIN_LED3);
  }
  else
  {
    ledOff(PIN_LED3);
  }
}

/**
 * @brief Activate GNSS device and start positioning.
 */
SDClass  theSD;
int take_picture_count = 0;


/**
 * Callback from Camera library when video frame is captured.
 */

void CamCB(CamImage img)
{

  /* Check the img instance is available or not. */

  if (img.isAvailable())
    {

      /* If you want RGB565 data, convert image data format to RGB565 */

      img.convertPixFormat(CAM_IMAGE_PIX_FMT_RGB565);

      /* You can use image data directly by using getImgSize() and getImgBuff().
       * for displaying image to a display, etc. */

      Serial.print("Image data size = ");
      Serial.print(img.getImgSize(), DEC);
      Serial.print(" , ");

      Serial.print("buff addr = ");
      Serial.print((unsigned long)img.getImgBuff(), HEX);
      Serial.println("");
    }
  else
    {
      Serial.print("Failed to get video stream image\n");
    }
}

/**
 * @brief Initialize camera
 */
//Motor A left motor
const int motorPin1  = 9;  // Pin 14 of L293
const int motorPin2  = 10;  // Pin 10 of L293
//Motor B right motor
const int motorPin3  = 8; // Pin  7 of L293
const int motorPin4  = 7;  // Pin  2 of L293
//Arm Motor 1
const int motorPin5  = 5;  
const int motorPin6  = 6;  
//Arm Motor 2
const int motorPin7  = 3;  
const int motorPin8  = 4;  
char val;
// defines pins numbers for ultrasonic sensor
const int trigPin = 12;
const int echoPin = 13;
// defines variables for ultrasonic sensor
long duration;
int distance;
void setup() {  
  
  /* put your setup code here, to run once: */

  int error_flag = 0;

  /* Wait HW initialization done. */
  sleep(3);

  /* Turn on all LED:Setup start. */
  ledOn(PIN_LED0);
  ledOn(PIN_LED1);
  ledOn(PIN_LED2);
  ledOn(PIN_LED3);

  /* Set Debug mode to Info */
  Gnss.setDebugMode(PrintInfo);

  int result;

  /* Activate GNSS device */
  result = Gnss.begin();

  if (result != 0)
  {
    Serial.println("Gnss begin error!!");
    error_flag = 1;
  }
  else
  {
    /* Setup GNSS
     *  It is possible to setup up to two GNSS satellites systems.
     *  Depending on your location you can improve your accuracy by selecting different GNSS system than the GPS system.
     *  See: https://developer.sony.com/develop/spresense/developer-tools/get-started-using-nuttx/nuttx-developer-guide#_gnss
     *  for detailed information.
    */
    switch (satType)
    {
    case eSatGps:
      Gnss.select(GPS);
      break;

    case eSatGpsSbas:
      Gnss.select(GPS);
      Gnss.select(SBAS);
      break;

    case eSatGlonass:
      Gnss.select(GLONASS);
      break;

    case eSatGpsGlonass:
      Gnss.select(GPS);
      Gnss.select(GLONASS);
      break;

    case eSatGpsQz1c:
      Gnss.select(GPS);
      Gnss.select(QZ_L1CA);
      break;

    case eSatGpsQz1cQz1S:
      Gnss.select(GPS);
      Gnss.select(QZ_L1CA);
      Gnss.select(QZ_L1S);
      break;

    case eSatGpsGlonassQz1c:
    default:
      Gnss.select(GPS);
      Gnss.select(GLONASS);
      Gnss.select(QZ_L1CA);
      break;
    }


    /* Start positioning */
    result = Gnss.start(COLD_START);
    if (result != 0)
    {
      Serial.println("Gnss start error!!");
      error_flag = 1;
    }
    else
    {
      Serial.println("Gnss setup OK");
    }
  }

  /* Turn off all LED:Setup done. */
  ledOff(PIN_LED0);
  ledOff(PIN_LED1);
  ledOff(PIN_LED2);
  ledOff(PIN_LED3);

  /* Set error LED. */
  if (error_flag == 1)
  {
    Led_isError(true);
    exit(0);
  }
    /* Open serial communications and wait for port to open */

  Serial.begin(BAUDRATE);
  while (!Serial)
    {
      ; /* wait for serial port to connect. Needed for native USB port only */
    }


  /* begin() without parameters means that
   * number of buffers = 1, 30FPS, QVGA, YUV 4:2:2 format */

  Serial.println("Prepare camera");
  theCamera.begin();


  /* Start video stream.
   * If received video stream data from camera device,
   *  camera library call CamCB.
   */

  Serial.println("Start streaming");
  theCamera.startStreaming(true, CamCB);

  /* Auto white balance configuration */

  Serial.println("Set Auto white balance parameter");
  theCamera.setAutoWhiteBalanceMode(CAM_WHITE_BALANCE_DAYLIGHT);
 
  /* Set parameters about still picture.
   * In the following case, QUADVGA and JPEG.
   */

  Serial.println("Start streaming");
  theCamera.setStillPictureImageFormat(
     CAM_IMGSIZE_QUADVGA_H,
     CAM_IMGSIZE_QUADVGA_V,
     CAM_IMAGE_PIX_FMT_JPG);
    pinMode(motorPin1, OUTPUT);
    pinMode(motorPin2, OUTPUT);
    pinMode(motorPin3, OUTPUT);
    pinMode(motorPin4, OUTPUT);
    pinMode(motorPin5, OUTPUT);
    pinMode(motorPin6, OUTPUT);
    pinMode(motorPin7, OUTPUT);
    pinMode(motorPin8, OUTPUT);
    pinMode(trigPin, OUTPUT); // Sets the trigPin as an Output
    pinMode(echoPin, INPUT); // Sets the echoPin as an Input
  // Serial.begin(9600);

}

/**
 * @brief %Print position information.
 */
static void print_pos(SpNavData *pNavData)
{
  char StringBuffer[STRING_BUFFER_SIZE];

  /* print time */
  snprintf(StringBuffer, STRING_BUFFER_SIZE, "%04d/%02d/%02d ", pNavData->time.year, pNavData->time.month, pNavData->time.day);
  Serial.print(StringBuffer);

  snprintf(StringBuffer, STRING_BUFFER_SIZE, "%02d:%02d:%02d.%06d, ", pNavData->time.hour, pNavData->time.minute, pNavData->time.sec, pNavData->time.usec);
  Serial.print(StringBuffer);

  /* print satellites count */
  snprintf(StringBuffer, STRING_BUFFER_SIZE, "numSat:%2d, ", pNavData->numSatellites);
  Serial.print(StringBuffer);

  /* print position data */
  if (pNavData->posFixMode == FixInvalid)
  {
    Serial.print("No-Fix, ");
  }
  else
  {
    Serial.print("Fix, ");
  }
  if (pNavData->posDataExist == 0)
  {
    Serial.print("No Position");
  }
  else
  {
    Serial.print("Lat=");
    Serial.print(pNavData->latitude, 6);
    Serial.print(", Lon=");
    Serial.print(pNavData->longitude, 6);
  }

  Serial.println("");
}

/**
 * @brief %Print satellite condition.
 */
static void print_condition(SpNavData *pNavData)
{
  char StringBuffer[STRING_BUFFER_SIZE];
  unsigned long cnt;

  /* Print satellite count. */
  snprintf(StringBuffer, STRING_BUFFER_SIZE, "numSatellites:%2d\n", pNavData->numSatellites);
  Serial.print(StringBuffer);

  for (cnt = 0; cnt < pNavData->numSatellites; cnt++)
  {
    const char *pType = "---";
    SpSatelliteType sattype = pNavData->getSatelliteType(cnt);

    /* Get satellite type. */
    /* Keep it to three letters. */
    switch (sattype)
    {
      case GPS:
        pType = "GPS";
        break;

      case GLONASS:
        pType = "GLN";
        break;

      case QZ_L1CA:
        pType = "QCA";
        break;

      case SBAS:
        pType = "SBA";
        break;

      case QZ_L1S:
        pType = "Q1S";
        break;

      default:
        pType = "UKN";
        break;
    }

    /* Get print conditions. */
    unsigned long Id  = pNavData->getSatelliteId(cnt);
    unsigned long Elv = pNavData->getSatelliteElevation(cnt);
    unsigned long Azm = pNavData->getSatelliteAzimuth(cnt);
    float sigLevel = pNavData->getSatelliteSignalLevel(cnt);

    /* Print satellite condition. */
    snprintf(StringBuffer, STRING_BUFFER_SIZE, "[%2d] Type:%s, Id:%2d, Elv:%2d, Azm:%3d, CN0:", cnt, pType, Id, Elv, Azm );
    Serial.print(StringBuffer);
    Serial.println(sigLevel, 6);
  }
}

/**
 * @brief %Print position information and satellite condition.
 *
 * @details When the loop count reaches the RESTART_CYCLE value, GNSS device is
 *          restarted.
 */

void loop() {
  static int LoopCount = 0;
  static int LastPrintMin = 0;

  /* Blink LED. */
  Led_isActive();

  /* Check update. */
  if (Gnss.waitUpdate(-1))
  {
    /* Get NaviData. */
    SpNavData NavData;
    Gnss.getNavData(&NavData);

    /* Set posfix LED. */
    bool LedSet = (NavData.posDataExist && (NavData.posFixMode != FixInvalid));
    Led_isPosfix(LedSet);

    /* Print satellite information every minute. */
    if (NavData.time.minute != LastPrintMin)
    {
      print_condition(&NavData);
      LastPrintMin = NavData.time.minute;
    }

    /* Print position information. */
    print_pos(&NavData);
  }
  else
  {
    /* Not update. */
    Serial.println("data not update");
  }

  /* Check loop count. */
  LoopCount++;
  if (LoopCount >= RESTART_CYCLE)
  {
    int error_flag = 0;

    /* Turn off LED0 */
    ledOff(PIN_LED0);

    /* Set posfix LED. */
    Led_isPosfix(false);

    /* Restart GNSS. */
    if (Gnss.stop() != 0)
    {
      Serial.println("Gnss stop error!!");
      error_flag = 1;
    }
    else if (Gnss.end() != 0)
    {
      Serial.println("Gnss end error!!");
      error_flag = 1;
    }
    else
    {
      Serial.println("Gnss stop OK.");
    }

    if (Gnss.begin() != 0)
    {
      Serial.println("Gnss begin error!!");
      error_flag = 1;
    }
    else if (Gnss.start(HOT_START) != 0)
    {
      Serial.println("Gnss start error!!");
      error_flag = 1;
    }
    else
    {
      Serial.println("Gnss restart OK.");
    }

    LoopCount = 0;

    /* Set error LED. */
    if (error_flag == 1)
    {
      Led_isError(true);
      exit(0);
    }
  }
digitalWrite(trigPin, LOW);
delayMicroseconds(2);
// Sets the trigPin on HIGH state for 10 micro seconds
digitalWrite(trigPin, HIGH);
delayMicroseconds(10);
digitalWrite(trigPin, LOW);
// Reads the echoPin, returns the sound wave travel time in microseconds
duration = pulseIn(echoPin, HIGH);
// Calculating the distance
distance= duration*0.034/2;
delay(500);
// Prints the distance on the Serial Monitor
Serial.print("Distance: ");
Serial.println(distance);  
  if (Serial.available())
  {
    val = Serial.read();
  delay(100);
if (val == 's')
{
//Backward code
   // Set Motor A backward
    digitalWrite(motorPin1, HIGH);
    digitalWrite(motorPin2, LOW);
    // Set Motor B backward
    digitalWrite(motorPin3, HIGH);
    digitalWrite(motorPin4, LOW);
    Serial.println("Backward");
    
    
}
if (val == 'w')
{
//forward code
    // Set Motor A forward
    digitalWrite(motorPin1, LOW);
    digitalWrite(motorPin2, HIGH);
    // Set Motor B forward
    digitalWrite(motorPin3, LOW);
    digitalWrite(motorPin4, HIGH);
  //  delay(2000);
}
if (val == 'a')
{    
//left code
    // Set Motor A backward
    digitalWrite(motorPin1, HIGH);
    digitalWrite(motorPin2, LOW);
    // Set Motor B forward
    digitalWrite(motorPin3, LOW);
    digitalWrite(motorPin4, HIGH);
    //delay(2000);
}
if (val == 'd')
{
//right code 
    // Set Motor A forward
    digitalWrite(motorPin1, LOW);
    digitalWrite(motorPin2, HIGH);
     // Set Motor B backward
    digitalWrite(motorPin3, HIGH);
    digitalWrite(motorPin4, LOW);
    //delay(2000);
  }
  if (val == 'i')
{
//Arm motor 1 operation 1
    digitalWrite(motorPin5, LOW);
    digitalWrite(motorPin6, HIGH);
  }
  if (val == 'k')
{
//Arm motor 1 operation 2
    digitalWrite(motorPin5, HIGH);
    digitalWrite(motorPin6, LOW);
  }
  if (val == 'j')
{
//Arm motor 2 operation 1
    digitalWrite(motorPin7, HIGH);
    digitalWrite(motorPin8, LOW);
  }
  if (val == 'l')
{
//Arm motor 2 operation 2
    digitalWrite(motorPin7, LOW);
    digitalWrite(motorPin8, HIGH);
  }
  if (val == 'c')
{
  sleep(1); /* wait for one second to take still picture. */

  /* You can change the format of still picture at here also, if you want. */

  /* theCamera.setStillPictureImageFormat(
   *   CAM_IMGSIZE_HD_H,
   *   CAM_IMGSIZE_HD_V,
   *   CAM_IMAGE_PIX_FMT_JPG);
   */

  /* This sample code can take 100 pictures in every one second from starting. */

  if (take_picture_count < 100)
    {

      /* Take still picture.
      * Unlike video stream(startStreaming) , this API wait to receive image data
      *  from camera device.
      */
  
      Serial.println("call takePicture()");
      CamImage img = theCamera.takePicture();

      /* Check availability of the img instance. */
      /* If any error was occured, the img is not available. */

      if (img.isAvailable())
        {
          /* Create file name */
    
          char filename[16] = {0};
          sprintf(filename, "PICT%03d.JPG", take_picture_count);
    
          Serial.print("Save taken picture as ");
          Serial.print(filename);
          Serial.println("");

          /* Save to SD card as the finename */
    
          File myFile = theSD.open(filename, FILE_WRITE);
          myFile.write(img.getImgBuff(), img.getImgSize());
          myFile.close();
        }

      take_picture_count++;
    }
}
  }
  else 
 {
  Serial.println("Serial not available");
    digitalWrite(motorPin1, LOW);
    digitalWrite(motorPin2, LOW);
    digitalWrite(motorPin3, LOW);
    digitalWrite(motorPin4, LOW);
    digitalWrite(motorPin5, LOW);
    digitalWrite(motorPin6, LOW);
    digitalWrite(motorPin7, LOW);
    digitalWrite(motorPin8, LOW);
 }
}

Image Processing Code

Python
# Copyright (c) 2017 Sony Corporation. All Rights Reserved.
#
# Licensed under the Apache License, Version 2.0 (the "License");
# you may not use this file except in compliance with the License.
# You may obtain a copy of the License at
#
#     http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing, software
# distributed under the License is distributed on an "AS IS" BASIS,
# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
# See the License for the specific language governing permissions and
# limitations under the License.

import nnabla as nn
import nnabla.logger as logger
import nnabla.functions as F
import nnabla.parametric_functions as PF
import nnabla.solver as S
import nnabla.utils.save as save

from args import get_args
from mnist_data import data_iterator_mnist

import os


def categorical_error(pred, label):
    """
    Compute categorical error given score vectors and labels as
    numpy.ndarray.
    """
    pred_label = pred.argmax(1)
    return (pred_label != label.flat).mean()


# Binary Connect Model
def mnist_binary_connect_lenet_prediction(image, test=False):
    """
    Construct LeNet for MNIST (BinaryNet version).
    """
    with nn.parameter_scope("conv1"):
        c1 = PF.binary_connect_convolution(image, 16, (5, 5))
        c1 = PF.batch_normalization(c1, batch_stat=not test)
        c1 = F.elu(F.average_pooling(c1, (2, 2)))
    with nn.parameter_scope("conv2"):
        c2 = PF.binary_connect_convolution(c1, 16, (5, 5))
        c2 = PF.batch_normalization(c2, batch_stat=not test)
        c2 = F.elu(F.average_pooling(c2, (2, 2)))
    with nn.parameter_scope("fc3"):
        c3 = PF.binary_connect_affine(c2, 50)
        c3 = PF.batch_normalization(c3, batch_stat=not test)
        c3 = F.elu(c3)
    with nn.parameter_scope("fc4"):
        c4 = PF.binary_connect_affine(c3, 10)
        c4 = PF.batch_normalization(c4, batch_stat=not test)
    return c4


def mnist_binary_connect_resnet_prediction(image, test=False):
    """
    Construct ResNet for MNIST (BinaryNet version).
    """
    def bn(x):
        return PF.batch_normalization(x, batch_stat=not test)

    def res_unit(x, scope):
        C = x.shape[1]
        with nn.parameter_scope(scope):
            with nn.parameter_scope('conv1'):
                h = F.elu(bn(PF.binary_connect_convolution(
                    x, C / 2, (1, 1), with_bias=False)))
            with nn.parameter_scope('conv2'):
                h = F.elu(
                    bn(PF.binary_connect_convolution(h, C / 2, (3, 3), pad=(1, 1), with_bias=False)))
            with nn.parameter_scope('conv3'):
                h = bn(PF.binary_connect_convolution(
                    h, C, (1, 1), with_bias=False))
        return F.elu(x + h)
    # Conv1 --> 64 x 32 x 32
    with nn.parameter_scope("conv1"):
        c1 = F.elu(
            bn(PF.binary_connect_convolution(image, 64, (3, 3), pad=(3, 3), with_bias=False)))
    # Conv2 --> 64 x 16 x 16
    c2 = F.max_pooling(res_unit(c1, "conv2"), (2, 2))
    # Conv3 --> 64 x 8 x 8
    c3 = F.max_pooling(res_unit(c2, "conv3"), (2, 2))
    # Conv4 --> 64 x 8 x 8
    c4 = res_unit(c3, "conv4")
    # Conv5 --> 64 x 4 x 4
    c5 = F.max_pooling(res_unit(c4, "conv5"), (2, 2))
    # Conv5 --> 64 x 4 x 4
    c6 = res_unit(c5, "conv6")
    pl = F.average_pooling(c6, (4, 4))
    with nn.parameter_scope("classifier"):
        y = bn(PF.binary_connect_affine(pl, 10))
    return y

# Binary Net Model


def mnist_binary_net_lenet_prediction(image, test=False):
    """
    Construct LeNet for MNIST (BinaryNet version).
    """
    with nn.parameter_scope("conv1"):
        c1 = PF.binary_connect_convolution(image, 16, (5, 5))
        c1 = PF.batch_normalization(c1, batch_stat=not test)
        c1 = F.binary_tanh(F.average_pooling(c1, (2, 2)))
    with nn.parameter_scope("conv2"):
        c2 = PF.binary_connect_convolution(c1, 16, (5, 5))
        c2 = PF.batch_normalization(c2, batch_stat=not test)
        c2 = F.binary_tanh(F.average_pooling(c2, (2, 2)))
    with nn.parameter_scope("fc3"):
        c3 = PF.binary_connect_affine(c2, 50)
        c3 = PF.batch_normalization(c3, batch_stat=not test)
        c3 = F.binary_tanh(c3)
    with nn.parameter_scope("fc4"):
        c4 = PF.binary_connect_affine(c3, 10)
        c4 = PF.batch_normalization(c4, batch_stat=not test)
    return c4


def mnist_binary_net_resnet_prediction(image, test=False):
    """
    Construct ResNet for MNIST (BinaryNet version).
    """
    def bn(x):
        return PF.batch_normalization(x, batch_stat=not test)

    def res_unit(x, scope):
        C = x.shape[1]
        with nn.parameter_scope(scope):
            with nn.parameter_scope('conv1'):
                h = F.binary_tanh(bn(PF.binary_connect_convolution(
                    x, C / 2, (1, 1), with_bias=False)))
            with nn.parameter_scope('conv2'):
                h = F.binary_tanh(
                    bn(PF.binary_connect_convolution(h, C / 2, (3, 3), pad=(1, 1), with_bias=False)))
            with nn.parameter_scope('conv3'):
                h = bn(PF.binary_connect_convolution(
                    h, C, (1, 1), with_bias=False))
        return F.binary_tanh(x + h)
    # Conv1 --> 64 x 32 x 32
    with nn.parameter_scope("conv1"):
        c1 = F.binary_tanh(
            bn(PF.binary_connect_convolution(image, 64, (3, 3), pad=(3, 3), with_bias=False)))
    # Conv2 --> 64 x 16 x 16
    c2 = F.max_pooling(res_unit(c1, "conv2"), (2, 2))
    # Conv3 --> 64 x 8 x 8
    c3 = F.max_pooling(res_unit(c2, "conv3"), (2, 2))
    # Conv4 --> 64 x 8 x 8
    c4 = res_unit(c3, "conv4")
    # Conv5 --> 64 x 4 x 4
    c5 = F.max_pooling(res_unit(c4, "conv5"), (2, 2))
    # Conv5 --> 64 x 4 x 4
    c6 = res_unit(c5, "conv6")
    pl = F.average_pooling(c6, (4, 4))
    with nn.parameter_scope("classifier"):
        y = bn(PF.binary_connect_affine(pl, 10))
    return y

# Binary Weight Model


def mnist_binary_weight_lenet_prediction(image, test=False):
    """
    Construct LeNet for MNIST (Binary Weight Network version).
    """
    with nn.parameter_scope("conv1"):
        c1 = PF.binary_weight_convolution(image, 16, (5, 5))
        c1 = F.elu(F.average_pooling(c1, (2, 2)))
    with nn.parameter_scope("conv2"):
        c2 = PF.binary_weight_convolution(c1, 16, (5, 5))
        c2 = F.elu(F.average_pooling(c2, (2, 2)))
    with nn.parameter_scope("fc3"):
        c3 = F.elu(PF.binary_weight_affine(c2, 50))
    with nn.parameter_scope("fc4"):
        c4 = PF.binary_weight_affine(c3, 10)
    return c4


def mnist_binary_weight_resnet_prediction(image, test=False):
    """
    Construct ResNet for MNIST (Binary Weight Network version).
    """
    def bn(x):
        return PF.batch_normalization(x, batch_stat=not test)

    def res_unit(x, scope):
        C = x.shape[1]
        with nn.parameter_scope(scope):
            with nn.parameter_scope('conv1'):
                h = F.elu(bn(PF.binary_weight_convolution(
                    x, C / 2, (1, 1), with_bias=False)))
            with nn.parameter_scope('conv2'):
                h = F.elu(
                    bn(PF.binary_weight_convolution(h, C / 2, (3, 3), pad=(1, 1), with_bias=False)))
            with nn.parameter_scope('conv3'):
                h = bn(PF.binary_weight_convolution(
                    h, C, (1, 1), with_bias=False))
        return F.elu(x + h)
    # Conv1 --> 64 x 32 x 32
    with nn.parameter_scope("conv1"):
        c1 = F.elu(
            bn(PF.binary_weight_convolution(image, 64, (3, 3), pad=(3, 3), with_bias=False)))
    # Conv2 --> 64 x 16 x 16
    c2 = F.max_pooling(res_unit(c1, "conv2"), (2, 2))
    # Conv3 --> 64 x 8 x 8
    c3 = F.max_pooling(res_unit(c2, "conv3"), (2, 2))
    # Conv4 --> 64 x 8 x 8
    c4 = res_unit(c3, "conv4")
    # Conv5 --> 64 x 4 x 4
    c5 = F.max_pooling(res_unit(c4, "conv5"), (2, 2))
    # Conv5 --> 64 x 4 x 4
    c6 = res_unit(c5, "conv6")
    pl = F.average_pooling(c6, (4, 4))
    with nn.parameter_scope("classifier"):
        y = PF.binary_weight_affine(pl, 10)
    return y


def train():
    """
    Main script.

    Steps:

    * Parse command line arguments.
    * Specify a context for computation.
    * Initialize DataIterator for MNIST.
    * Construct a computation graph for training and validation.
    * Initialize a solver and set parameter variables to it.
    * Create monitor instances for saving and displaying training stats.
    * Training loop
      * Computate error rate for validation data (periodically)
      * Get a next minibatch.
      * Set parameter gradients zero
      * Execute forwardprop on the training graph.
      * Execute backprop.
      * Solver updates parameters by using gradients computed by backprop.
      * Compute training error
    """
    args = get_args(monitor_path='tmp.monitor.bnn')

    # Get context.
    from nnabla.ext_utils import get_extension_context
    logger.info("Running in %s" % args.context)
    ctx = get_extension_context(
        args.context, device_id=args.device_id, type_config=args.type_config)
    nn.set_default_context(ctx)

    # Initialize DataIterator for MNIST.
    data = data_iterator_mnist(args.batch_size, True)
    vdata = data_iterator_mnist(args.batch_size, False)

    # Create CNN network for both training and testing.
    mnist_cnn_prediction = mnist_binary_connect_lenet_prediction
    if args.net == 'bincon':
        mnist_cnn_prediction = mnist_binary_connect_lenet_prediction
    elif args.net == 'binnet':
        mnist_cnn_prediction = mnist_binary_net_lenet_prediction
    elif args.net == 'bwn':
        mnist_cnn_prediction = mnist_binary_weight_lenet_prediction
    elif args.net == 'bincon_resnet':
        mnist_cnn_prediction = mnist_binary_connect_resnet_prediction
    elif args.net == 'binnet_resnet':
        mnist_cnn_prediction = mnist_binary_net_resnet_prediction
    elif args.net == 'bwn_resnet':
        mnist_cnn_prediction = mnist_binary_weight_resnet_prediction

    # TRAIN
    # Create input variables.
    image = nn.Variable([args.batch_size, 1, 28, 28])
    label = nn.Variable([args.batch_size, 1])
    # Create prediction graph.
    pred = mnist_cnn_prediction(image / 255, test=False)
    pred.persistent = True
    # Create loss function.
    loss = F.mean(F.softmax_cross_entropy(pred, label))

    # TEST
    # Create input variables.
    vimage = nn.Variable([args.batch_size, 1, 28, 28])
    vlabel = nn.Variable([args.batch_size, 1])
    # Create prediction graph.
    vpred = mnist_cnn_prediction(vimage / 255, test=True)

    # Create Solver.
    solver = S.Adam(args.learning_rate)
    solver.set_parameters(nn.get_parameters())

    # Create monitor.
    import nnabla.monitor as M
    monitor = M.Monitor(args.monitor_path)
    monitor_loss = M.MonitorSeries("Training loss", monitor, interval=10)
    monitor_err = M.MonitorSeries("Training error", monitor, interval=10)
    monitor_time = M.MonitorTimeElapsed("Training time", monitor, interval=100)
    monitor_verr = M.MonitorSeries("Test error", monitor, interval=10)

    # Training loop.
    for i in range(args.max_iter):
        if i % args.val_interval == 0:
            # Validation
            ve = 0.0
            for j in range(args.val_iter):
                vimage.d, vlabel.d = vdata.next()
                vpred.forward(clear_buffer=True)
                ve += categorical_error(vpred.d, vlabel.d)
            monitor_verr.add(i, ve / args.val_iter)
        if i % args.model_save_interval == 0:
            nn.save_parameters(os.path.join(
                args.model_save_path, 'params_%06d.h5' % i))
        # Training forward
        image.d, label.d = data.next()
        solver.zero_grad()
        loss.forward(clear_no_need_grad=True)
        # Training backward & update
        loss.backward(clear_buffer=True)
        solver.weight_decay(args.weight_decay)
        solver.update()
        # Monitor
        e = categorical_error(pred.d, label.d)
        monitor_loss.add(i, loss.d.copy())
        monitor_err.add(i, e)
        monitor_time.add(i)

    parameter_file = os.path.join(
        args.model_save_path, 'params_%06d.h5' % args.max_iter)
    nn.save_parameters(parameter_file)


if __name__ == '__main__':
    train()

Credits

VARUN SHARMAN 1660484

VARUN SHARMAN 1660484

3 projects • 8 followers
A thinker, developer, and designer; continually re-imagining what’s possible to help bring in a change in the world.
Jival Jenson

Jival Jenson

0 projects • 5 followers

Comments