Some of our projects
TimBer measurement using 2d approach
#YOLOv5, #OpenCV, #Python
Challenge
VMF, a leading wood measurement service provider in Latvia, required an efficient and accurate stack measurement solution.
Solution
Our team has quickly developed an AI driven system using comprehensive data analysis from over 12 000 photos.
The system, trained to recognize patterns and dimensions, provided precise measurements even in challenging lighting and weather conditions.
Result
• Reduced measurement operational costs
• 99% accuracy within 2 seconds per image
• 24/7 service availability scalable to meet varying demand
Ai generated product labels
#StableDiffusion, #OpenCV, #Python
Challenge
Supliful, a prominent player in the print-on-demand dropshipping industry, has sought our expertise to develop a user-friendly solution for seamless product label design creation.
Solution
Our team has quickly came up with the proof-of-concept of the system allowing to create product label backgrounds using open-source generative AI model.
The research included novel cost-effective approach to accelerate generation process using background tiles, which made user experience almost realtime.
Additionally, we came up with the list of friendly UX controls for easier prompt selection, and effectively automated text color adaptation to the background for enhanced visibility.
Result
• Accelerated onboarding for new users
• Enhanced design generation speed
• Improved overall user experience
Timber measurement using 3d digital twins
#Open3D, #Python, #Potree
Challenge
KPDC aimed to improve the accuracy and speed of freight volume measurement at their checkpoints.
Solution
Our team created a point cloud solution using 3D scanning to accurately model each trailer and cargo for precise load measurement.
The technology not only accurately tracks the volume of timber but also identifies license plates and counts the number of trailers on the go.
Result
• Reduced operational costs per measurement
• Improved service availability and accuracy
• Integrated with client IT accounting systems
Automated face recognition
#DeepFace, #VGGFace #Python
Challenge
Large-scale events such as marathons, concerts, or conferences typically generate photo galleries with thousands of images.
Traditional methods of manually searching through these photos are slow and ineffective leading to frustration and compromised user experience.
Solution
Photolab’s face recognition solution offers an instant search capability for thousands of online photos, with integration for effortless sharing on social media and email.
Result
• 10 Client branded sites published with 58 event galleries
• 8000+ searches more than 40 000 photos in search results
• 3000+ users linked their Facebook/Instagram account to our system