BlogCases

Our team build an automation for the marketplace app

Cases

Marketplace for selling and buying physical products.

Business Task

The client needed to automate and simplify uploading new items for selling by implementing noise-resistant automatic categorization of the uploaded products and reliable price prediction to enhance the UX.

Solutions

Our team has built the computer vision solution to classify objects by photos. Our team improved the object detection and classification algorithm to be noise-resistant to images of the product that have other objects. Our team built a clustering algorithm specific to the customer's category hierarchy that always changes due to different reasons. This pipeline was completely automated and all the algorithms ran periodically to make sure all the data is up to date on the marketplace's site.

Value delivered by Dysnix

Our team owned research and development of ML/AI part of the product. We've significantly simplified the process of the product upload in the marketplace mobile apps. This increased user satisfaction and user retention.

Conclusion

Our engineers has been involved in the development of the AI algorithms to increase user retention of the marketplace. Later on we've been owning the AI part of the product and continue to build strategy of the AI part as well as development and improvement of the existing algorithms.

Industry

Data Science

Computer vision

Location
USA
Project duration
2019
Our team
4 (1 CTO, 2 Computer Vision Engineers, 1 Full-Stack Developer)
Technologies used

TensorFlow

Keras

PyTorch

NumPy

RetinaNet

Cascade RCNN

VGG

Inception

Docker

Services provided

AWS