An ML based solution for photo processing

Type Machine learning algorithm, Web application
TECHNOLOGIES OpenCV, Python, TensorFlow
AREAS OF EXPERTISE Image processing, Machine Learning
TEAM 2 developers

Our client is an agency specialized in high-volume image processing. They already had their own image processing algorithms for extracting images from backgrounds. However, it didn’t work for all the types of backdrops they wanted to operate with. That is why they came to us with a clear task, to develop an ML based solution.

The main application area of the project is photo processing. One of the most common and sought-after tasks here is to replace the original backdrop of an image due to the need either to fix some issues with a background or to make the image more impressive and unique in general. It is also one of the most time-consuming and tedious tasks in the area of photographic processing. Doing this process manually will take an average specialist significantly more time than doing it automatically with an ML based solution. There’s plenty of solutions designed to help specialists with that task on the market built with Computer Vision mathematical algorithms and Machine Learning approach. However, the existing solutions proved to be either too slow for the real production needs or not precise enough.


Within this project, we've tested several state-of-the-art neural network architectures in application to the problem of accurate segmentation and computing the transparency levels for border pixels. This work is intended for processing photos with people captured in full length.


We've overcome several challenges connected with optimal data preprocessing necessary for training the network and processing the high-resolution images keeping their original size by making and aggregating predictions for overlapping patches.


The project comes up with the solution which helps photomontage specialists to handle high-volume tasks with the required speed and precision where 4000 x 3000-pixel images can be processed in roughly 5 seconds. The resulting solution operates with the following types of backdrops: 

  1. natural, 
  2. painted, 
  3. lastolite.

Business value

  • A twofold reduction in time spent by designers on photo processing;
  • The payback time for the project was 2 months.
An algorithm to recognize dead and live cells in yeast images
An algorithm to recognize dead and live cells in yeast images
Concrete crack detection algorithm
Concrete crack detection algorithm