---

This site uses cookies. Read more.

Computer Vision and Machine Learning Solutions

FAST DEVELOPMENT
1 WEEK
from concept
to working prototype
OBJECT DETECTION
SEE MORE
than human auditors
are capable of
IMAGE CLASSIFICATION
AUTOMATE
tasks that once
took many hours
SATELLITE IMAGE ANALYSIS
25cm
maximum satellite
image resolution

ML Frameworks and Custom Solutions

Appsilon Data Science builds machine learning models that can quickly identify objects, defects, and patterns that human auditors miss. We have a library of proprietary frameworks and workflows that allow for rapid prototyping and fast deployment. We are experts in:

  • Vehicle Damage Detection
  • Manufacturing Defect Detection
  • Fraud Detection
  • Optical Character Recognition
  • Natural Language Processing
  • Satellite Image Procurement and Analysis

We use state-of-the-art approaches and the latest research breakthroughs to deliver modern machine learning and computer vision solutions for NGOs, academic institutions, and enterprise clients.

Image
Honey Honey

Amplify Your Business
With Machine Learning

Our team can quickly build custom ML solutions or adapt existing solutions to your needs. Typically, we can build an ML PoC within one week. We have built models to classify wild animals in real-time to mitigate poaching, analyze satellite images to assess damage after natural disasters, and help businesses automatically detect defects in consumer products.

Case studies:

AI for Assisting Natural Disaster Recovery

Satellite Image Analysis fast.ai PyTorch

The Appsilon Data Science Machine Learning team recently took part in the xView2 competition organized by the Defense Innovation Unit (United States Department of Defense). Participants set out to utilize satellite imagery data to assist humanitarian efforts during natural disasters.

We were asked to build ML models using the novel xBD dataset provided by the organizers to estimate damage to infrastructure with the goal of reducing the amount of human labour and time required to plan an appropriate response.

Read more about the project here.

Interact with a demo of the final app here.

ML Wildlife Image Classification to Analyze Camera Trap Datasets

Realtime Image Classification Open Source Solution

We worked with biodiversity conservationists at the National Parks Agency in Gabon in collaboration with experts from the University of Stirling to build an ML model that automatically identifies wildlife from camera trap images. We completed a fully functional Computer Vision ML model in two weeks. Our model was able to identify animals in images that human auditors frequently missed. Our model can quickly and accurately process millions of camera trap images, saving thousands of human work hours. We received additional support for this project from the Google for Education fund.

Learn more about the project here.

Watch a video of the model in action here.

  • ML Model to Identify Wildlife 
  • Significantly More Accurate Than Human Auditors
  • Part of our AI For Good Initiative

Defect Detection

fast.ai PyTorch Starlette

We were approached by a major manufacturer to build a Machine Learning model for automatically detecting defects in cast iron products. We were able to train a working model and package it within an app for using the model as a prototype within two working days.

We used fast.ai on PyTorch (kaggle and GCP) for training the model. The app was built in Starlette.

  • 99.6% Test Accuracy for Identifying Defects
  • Two Days From Concept to Working Prototype
  • Reliable Solution for Quality Control

Appsilon proved to be an excellent business partner. We highly recommend working with them. 

John Dannberg, Principal, Boston Consulting Group

Appsilon proved to be an excellent business partner. We highly recommend working with them. 

John Dannberg,
Principal, Boston Consulting Group

Appsilon was able to quickly develop innovative ML solutions that we had not previously envisioned.

Robin Whytock, Researcher, University of Stirling

Appsilon was able to quickly develop innovative ML solutions that we had not previously envisioned.

Robin Whytock,
Researcher, University of Stirling
Marek Rogala
Marek Rogala
CTO

Reach out. We’ll respond within 24 hours.