ChakGo_Scan

Machine learning for tinkerhub build from home

Table of Contents

• Project name
• Team members
• Team id
• Link to product walkthrough
• How it works
• Libraries used
• How to configure
• How to run
• References

---

Project Name

ChakGo_Scan

ChakGo_Scan is a Machine learning model developed as part of Build From Home organized by Tinkerhub. This project is trained to distinguish between photo of Chakka or Manga ( Jackfruit or Mango). we came up with the name Chak(CHAKka)Go(manGO)_ Scan.We, a bunch of students have used tensorflow and keras for this ML code.

ChakGo_Scan is an easy to use platform for anyone who wants to check whether an image is Chakka or Manga. Simply open the Final_project.ipynb file and upload your picture

---

Team members

• Vaishakh v vaishakh-v
• Sudarsan R Mohan SUDARSAN-RM
• Darshan S darshanchaithram

---

Team id

• BFH/recaLYum338MTMIJQ/2021

---

Link to product walkthrough

https://drive.google.com/file/d/1LVdnxt5MZytjhrN0yYaHWaewYODOHmN8/view?usp=sharing

---

How it works

The project is divided into 3 colab files. The first one, namely 'Image dataset using selenium' is used to generate dateset from the internet and to scrape the images as per our criteria.

• We have used selenium framework to generate images.
• Parameters are passed as a CSV file.
• The downloaded images are scrapped as 224 x 224 and ensured to be RGB colourspace.
• We have also included a feature to add and scrape images from Google drive.
• All these images can be downloaded as a zip file.

The second file, namely 'Model training' is used to train our model with our dataset.

• Import dataset from Google drive.
• Grouped the images as training dataset and validation dataset.
• Data Augmentation technique is used to enlarge our dataset size.
• We imported the CNN model 'VGG16' from Tensorflow using keras library.
• Since this is a Binary classification, we need not train all the layers of 'VGG16'.
• Parameters are added to the model as per the criteria.
• Model is trained using our Augmented dataset with 10 Epochs.
• Trained model is saved into the Google drive.

The final file, namely 'Final Project' is where the user tests an image

• The model is imported from Google drive.
• The user can upload an image and predict if the image contain "Chakka" or "Manga".
• The results shows the uploaded image and the fruit type.

---

Libraries used

• Tensorflow - 2.4.1
• Matplotlib - 3.2.2
• tensorflow.keras
• Pandas
• PIL
• Selenium
• Requests
• Sys
• tensorflow.keras.preprocessing.image

---

How to configure

• Open the 'Final project.ipynb' in a jupyter notebook.
• Mount the google drive.

---

How to run

• Open the 'Final project.ipynb'
• Load the model from Google drive.
• Add a test image.
• Pass the image as a parameter to the model.
• Chakka/Manga/Sorry I cannot identify this image, as well as the upoaded picture will be displayed.

---

References

• towardsdatascience
• w3schools
• kaggle jackfruit
• kaggle mango
• Colab