Detection-of-pests

Asian University Machine Learning Camp Jeju 2018

Table of Contents

1. Overview
   • Datasets Used
2. Workflow
   • Introduction
   • Pre-requsites
   • Complete Workflow
   • Steps to Follow
3. References
4. Acknowledgement

Overview/Motivation

Agriculture is a most important and ancient occupation in India. As economy of India is based on agricultural production, utmost care of food production is necessary. Pests like virus, fungus and bacteria causes infection to plants with loss in quality and quantity production. There is large amount of loss of farmer in production. Hence proper care of plants is necessary for same. Image processing provides more efficient ways to detect diseases caused by fungus, bacteria or virus on plants. Mere observations by eyes to detect diseases are not accurate. Excess use also damages plants nutrient quality. It results in huge loss of production to farmer. Hence use of image processing techniques to detect and classify diseases in agricultural applications is helpful.

Datasets Used

For this project, the datasets are images: you can download images in 2 ways

1. Fatkun batch Download (chrome Extension)
2. FFmpeg - Command line tool (extracts images from videos)

Workflow

The workflow is divided into 4 main parts.

Introduction

ImageNet is a common academic data set in machine learning for training an image recognition system. Code in this directory demonstrates how to use TensorFlow to train and evaluate a type of convolutional neural network (CNN) on this data set.

http://arxiv.org/abs/1512.00567

This network achieves 21.2% top-1 and 5.6% top-5 error for single frame evaluation with a computational cost of 5 billion multiply-adds per inference and with using less than 25 million parameters. Below is a visualization of the model architecture.

Pre-Requsite

• Python 3.5+
• Tensorflow
• Docker Toolbox
• Android Studio

Complete Workflow

Why to use pre-trained model when we can build new one??

How does the pre-trained network works?

Steps to Follow

1. Install python 3.5 or above, tensorflow & Docker toolbox.

2. Place all your Datasets in a folder & The classification script uses the folder names as label names, and the images inside each folder should be pictures that correspond to that label. Collect as many pictures of each label as you can and try it out!

3. Clone the repository & navigate to the directory.

   git clone https://github.com/r-karthik/Detection-of-pests.git
   cd Detection-of-pests

4. For training you can download the sample flower Datasets & extract them to tf_files/training_dataset folder.

   └───training_dataset
      ├───Fruit Piercing Moth
      ├───Gall flies
      ├───Leaf-feeding caterpillars
      └───Scrab Beetle

5. Run the retrain.py in docker with the following command. As it trains, you can start TensorBoard in background. you'll see a series of step outputs, each one showing training accuracy, validation accuracy, and the cross entropy:

   • The training accuracy shows the percentage of the images used in the current training batch that were labeled with the correct class.
   • Validation accuracy: The validation accuracy is the precision (percentage of correctly-labelled images) on a randomly-selected group of images from a different set.
   • Cross entropy is a loss function that gives a glimpse into how well the learning process is progressing. (Lower numbers are better.)

       #setting up the architecture
       IMAGE_SIZE=224
       ARCHITECTURE="mobilenet_0.50_${IMAGE_SIZE}"
       #The default training is 4000 iterations.
       python -m scripts.retrain --bottleneck_dir=tf_files/bottlenecks --model_dir=tf_files/models/"${ARCHITECTURE}" --summaries_dir=tf_files/training_summaries/"${ARCHITECTURE}" --output_graph=tf_files/retrained_graph.pb --output_labels=tf_files/retrained_labels.txt --architecture="${ARCHITECTURE}" --image_dir=tf_files/training_dataset
       #TensorBoard
       tensorboard --logdir tf_files/training_summaries &

   The training completed with 92% accuracy..

6. when the training is completed the output is retrained_graph.pb & retrained_labels.txt you can test the model using the graph.

   #Testing
   python -m scripts.label_image --graph=tf_files/retrained_graph.pb --image=tf_files/testing_dataset/image_name.jpg

   

7. If you want to run the model to predict on android device you have to opimize the model some of the methods for compressing the model are:

   • Freeze the Graph : Converts Variables to Constants
   • Graph Transform Tool : Removes unnessary parts
   • Quantize Weights : Quantizes weights & calculations
   • quantize Calculations : Converts 32-bit Floating point calculations to 8-bit Integer by maintaining decent Accuracy.
   • Meamory Mapping : uses Meamory mapping to load parameters instead of standard File I/O.

8. You can perform the optimization to the retrained_graph.pb file so that the model can be compressible by running the following commands which can compress the model upto 70%.

   #optimized graph
    python -m tensorflow.python.tools.optimize_for_inference --input=tf_files/retrained_graph.pb --output=tf_files/optimized_graph.pb --input_names="input" --output_names="final_result"
   #rounded graph
   python -m scripts.quantize_graph --input=tf_files/optimized_graph.pb --output=tf_files/graph.pb --output_node_names=final_result --mode=weights_rounded

9. Paste the graph.pb & labels.txt in android/tfmobile/assets & open android studio choose existing project tfmobile. let the Gradle build finish & run your app. Now you can use your andoid mobile to classify things my using the App.

References

1. https://ieeexplore.ieee.org/document/7155951/

2. https://ieeexplore.ieee.org/document/7916653/

3. https://www.sciencedirect.com/science/article/pii/S0168169917311742?via%3Dihub

4. https://www.researchgate.net/publication/322941653_Deep_learning_models_for_plant_disease_detection_and_diagnosis

Acknowledgement

I would sincerely like to thank Jeju National University, Jeju Development Center & all the sponcers who have supported us for our work. And i would specially like to thank professor Yungcheol Byun for guiding us & helping us in every situation.