Looking for a developer familiar with Computer Vision (SIFT) with strong experience with Bag of Words (large scale data search)
We would like to build a prototype that will recognize labels on actual 3D products and provide an instant rating. Creation of an iphone app with supporting backend DB/server, to demonstrate the functionality specified below.
So far the image recognition is working but we need an expert of Visual Words and large scale date search to implement an appropriate index.
1. Use handheld camera for capture and transfer a low resolution file to server for speed performance. Product label format on products are fairly similar in sizing. Recognition calculation will happen on server, while resizing and pretreatment will happen locally.
2. Using the data retrieved, conduct a database lookup and accurately identify the product ID. Display “OK” Screen with DB rating and stored data (name of product, manufacturer and year).
If unable to identify a unique product (depending on accuracy of technology), provide a list of closest selection to choose. Display “OK level 2” Screen with DB rating and stored data (name of product, manufacturer and year).
Objective : Targeting a 75% accuracy recognition (“OK”) among product listed in DB (total of 60,000)
3. If name cannot be found provide a manual entry screen for manual search (same 3 criteria)
- If name is found in DB display “OK level 3” screen with rating and stored data (DB)
- If name cannot be found in DB (not a listed product) display “Sorry not listed” screen and entered data and picture
Objective: We are targeting a result screen to be provided within 10 seconds from capture.
4. “Learning Machine” logic
For products that cannot be identified (“Sorry not listed”), based upon validation of DB Administrator (visual confirmation of uploaded picture VS actual label) provide an interface to add this new product to the DB.
Objective: For a second capture of this product, we are looking for a positive identification.
Log all requests with a final status (OK, OK level 2, OK level 3, Sorry not listed) for further data analysis
1. SIFT Technology understanding (a working C++ engine will be provided)
2. Visual Words
3. Large scale data search
- Phonegap framework required (further HTML5 implementations).
- Linux Servers with PHP ideally & MySQL.
- Hosting on following PAAS Cloud Server.
- Simple wireframe UI without design implemented (ready for HTML 5 design integration)
- Simple web access to manage DB and add new products (testing purpose)
Payment upon receipt of the code. More details on request.
In your answer, please provide us rough information about your development strategy to meet our objectives. As well as details on your experience in that area.