#FoodAdvisor

FoodAdvisor is a simple food image search app that helps people to find where to eat with the given food image. It is also a Final Project for CS598 Visual Information Retrieval in Stevens Institute of Technology.

##Artechiture

In this project, We use Flask framework to create a RESTful service that serves the data for our Front-end rich client application written by AngularJS. Since we chose JSON as our API format, we chose MongoDB as our JSON data store.

##Technology Stacks

###Back-end

API Server: Flask 0.10.1

Persistence: MongoDB 2.6

Image Search: OpenCV-Python, Scipy Library, Numpy

Data source: Yelp(food images), Yelp API(business info)

###Front-end

App: AngularJS, Lo-Dash, Yeoman, Bower, Grunt

Style: Bootstrap, LESS, FontAwesome

##Development

###Folder Structure FoodAdvisor/ ├── app │ ├── apphelpers (Self-created global use functions) │   ├── dbhelpers (Self-created db-related functions) │   ├── imagesearchapis (Implements image search apis) │   ├── restapis (Implements RESTful apis ) │   ├── routes (App routes) │   └── static (Front-end Angular App) │ └── outputs (folder for saving output files) ├── flask (Flask virtual environment) ├── node_modules (Node dependencies)

###Configuration ####Environment

1. Install Python, MongoDB, OpenCV-Python.

2. Under the root directory, install python dependencies.

`$ ./installtolocal.sh`

3. Run the service(under the root directory), the server should be running without any issues.

`$ python run.py`

Note: If there is an error about nltk, try to install it manually as follow:

`$ python -m nltk.downloader stopwords`

####Data

• Run preparedata.py script under app/dbhelpers/ to insert data into database(Make sure your local mongodb server is openning).

  $ cd app/dbhelpers
  $ python preparedata.py
  

• Put image files into app/static/images/

• Put tfidf file into app/outputs/

####To Do List

#####Image Searching Algorithm:

• Collect images
• Extract SIFT features from all the images and save them to SIFT pool
• Cluster all SIFT features
• Re-represent each image with bag of words
• Calculate tf-idf for all images
• Do inverted file indexing
• Accept incoming image as query
• Return highest rank images

#####Back-end server:

• Text suggestion search from the field including ‘description’, ‘category’, ‘name’.
• Full text search from the field including ‘description’, ‘category’, ‘name’.
• REST API for searching food image with the given image file.
• REST API for Search food image with the given text: picture description, location or business name.

#####Front-end:

• A Flat, minimal looking interface.
• Single page.
• Text searching bar with autocomplete support.
• Upload file through XHR.
• Single upload file button with automatic submit.
• Menu button for sorting result.
• Load more button for pagination.
• Progress bar for any long wait operation.
• Animations for better user experience.

###RESTful Service ####Text autocomplete

####food image/text search

###DB document sample { "description": "Amazing chicken tikka and aloo tacos", "abspath": "/Users/hanyan/Desktop/Homework/CS598/FoodAdvisor/app/static/images/foods/23rd-street-cafe-los-angeles/Amazing chicken tikka and aloo tacos.jpg", "business_id": "23rd-street-cafe-los-angeles", "image_id": 0, "relpath": "images/foods/23rd-street-cafe-los-angeles/Amazing chicken tikka and aloo tacos.jpg", "business_info": { "category": [ [ "Indian", "indpak" ], [ "Mexican", "mexican" ], [ "American (New)", "newamerican" ] ], "rating": 4.0, "review_count": 178, "name": "23rd Street Cafe", "phone": "+1-213-749-1593", "location": { "display_name": [ "936 W 23rd St", "University Park", "Los Angeles, CA 90007" ], "details" : { "type" : "Point", "coordinates" : [ -118.2808464, 34.033785 ] } } } } ###Result sample { "result": [...], "status": {"text": last text query, "file": the file name(server side) of the last image query} }