Background Image Effects using PyTorch

When you miss the vacations and holidays during Covid-19, some people re-live their memories by seeing the old pictures. Now using the Tech, you can teleport yourself to any place virtually.

Take a selfie and move to another dimension of the world.

With the Mobile App & Artificial Intelligence (AI), You can change the background using another image, blur the background, remove the background, apply grayscale effect only on the background. The picture can be taken from Camera or select the image from Gallery!

Change Background

Take a selfie or group picture with family or friends and instantly change the background using your preferred place like Taj Mahal, Eiffel Tower, Golden Gate Bridge, or any of your preferred place.

Remove Background

You can remove the background for uploading photo for ID cards or any purpose where you need only the foreground.

Blur Background

Blur the background to create stunning effects on the images similar to photos taken using DSLR or professional camera.

Grayscale Background

You can apply Black & White or Grayscale effect only on the background thus creating amazing pictures with foreground picture in color and background in grayscale.

Technologies Used

Mobile App is developed using React Native. Image effects are achieved using Machine Learning frameworks: PyTorch, TorchVision and Python, Flask & OpenCV.

App Screenshots

Kids having fun with the App

AR & AI implementation in the Mobile App using BASNet

We always wish things to be alive and real. As a kid, we wanted superhero action figures to interact with us.

This Mobile App lets you extract the particular object from real life surroundings using Live Camera and move around the same object, even your superheroes!

Spider-Man in Action

Spider-Man Action using AI

BASNet Service

Boundary-Aware Salient Object Detection (BASNet) AI model is used for Salient detection and background removal. PyTorch & Torchvision libraries are required.

Image Processing Service

Python & Pillow (Python Image Library) is used for image processing and composite of the images.

Bumblebee (Transformers) in Action

Bumblebee (Transformers) Action using AI

Mobile App

React Native platform is used for creating Cross Platform Mobile Application Development for iOS & Android.

Thanks and Acknowledgements

http://openaccess.thecvf.com/content_CVPR_2019/html/Qin_BASNet_Boundary-Aware_Salient_Object_Detection_CVPR_2019_paper.html

https://github.com/cyrildiagne/ar-cutpaste

AI for Face tracking, face landmarks in Image or Video

Face tracking, bounding box around the face and 3D Face landmarks like lips, eyebrows, eyes, nose & cheek are identified with the Facemark application using Machine Learning model & Artificial Intelligence.

Facial landmarks are identified in the images, live videos taken from mobile phone front camera or computer webcam.

See below the screenshot of Demo App: