Describe in 250 words what the feature or service will do and how you’ll use Machine Learning to push the bar:
As a professional photographer my uncle spends around eight to ten hours in culling of 4000 photographs after every wedding shoot.
The world is moving towards mobile cameras from DSLR.
An on-device ML algorithm can automate the process to take down the time to 5 minutes or less cutting down 99% of the actual time spent in the process.
With on device machine learning, the service will be able to classify images into blurry, out of focus, overexposed, underexposed, duplicates and best shots.
This will provide a professional photographer with more time to grow his photography business, and a hobbyist to save a lot of time in figuring out her best shots.
Tell us how you plan on bringing it to life.
Describe where your project is, how you could use Google’s help in the endeavor, and how you plan on using On-Device ML technology to bring the concept to life. The best submissions have a great idea combined with a concrete path of where you plan on going, which should include:
- (1) any potential sample code you’ve already written,
- (2) a list of the ways you could use Google’s help,
- (3) as well as the timeline on how you plan on bringing it to life by May 1, 2020.
We are working on the primary market research of this concept from past 3-4 months and results are really convincing. For the prototype, we have used AutoML to classify photos into blur and unblur and overexposed and underexposed with an accuracy of 81% and 82% respectively.
With the help of Google and the best datasets, we are aiming to achieve a higher accuracy in our algorithms along with a possible integration of this in Google Photos.
Demo link: https://drive.google.com/file/d/1tt2JZPNd1QyShZr9fxt7YF9SxoIxByCO/view
With the help of AutoML and our proprietary algorithms, we have already created a Proof of concept.Timeline:
Dec 2 - Dec 9 : Creating the user interface
Dec 9 - Dec 15: Releasing Proof of Concept with blurry, out of focus and duplicates classification
Dec 15 - Dec 22: Creating datasets, training models on Blinks, emotions, cropped and sharpness on AutoML
Dec 22- Dec 29: Testing Models and Evaluating Results
Dec 29 - Jan 6 : Integrating New Models with User Interface
Jan 6 - Jan 12: Testing the Integration and handing over to users
Jan 12- Jan 19: Asking Feedback to reiterate
Jan 19 - Feb 1 Working on Iteration
Feb 1 - Feb 7: Releasing MVP
Feb 7 - Feb 14: Work on Algorithm Accuracy
Feb 14 - Feb 21: Testing and integrating better accuracy models
Feb 21 - Mar 1: Customer Feedback
Mar 1 - Mar 8 : Reiterating on Customer Feedback
Mar 9 - Mar 16: Launch
A great idea is just one part of the equation; we also want to learn a bit more about you. Share with us some of your other projects so we can get an idea of how we can assist you with your project.
I am a Google Developer Expert for Web and Angular and an avid blogger. Being a ML facilitator for Google, I have grown the Google AI Community India from scratch upto 700 members today. After nominated as the first guest author on Google Codelabs some of my codelabs for ML and Angular has helped the ML community. My experience lies in solving real world problems by learning new technologies quickly.
Projects that I have worked on:
- Open-event-wsgen : A platform to generate websites automatically for events, meetups, and conferences.
- Roobits: An GCP infrastructure to collect and analyse data in real-time.
- Motley: A UI framework to follow the ITCSS design language
- Eye-bot: A R-PI based IOT device control the prosthetic on eye blinks.
Popular Blog posts:
Using Genetic Algorithms to Automate Chrome Dino Game