Computer Vision Expertise

Example applications

Computer Vision spans a wide range of techniques for analysing and understanding images in various forms. The examples below are only a snapshot of the many possible applications: from simple object recognition to fine-grained interpretation of complex scenes.

Object detection consists in automatically recognising and locating an element within an image. This often provides more precision than classification alone, by generating a bounding box around the detected object. Having this visual localisation can be crucial for use cases like video surveillance or identifying products in-store.

Semantic segmentation automatically assigns a category—or mask—to every pixel in an image, precisely delineating different areas. Segmentation models are particularly relevant for domains such as high-precision mapping for navigation systems or detailed tissue analysis in healthcare.

Pose estimation analyses and locates key points of the human body, such as joints, to establish posture and infer movement. This approach is crucial in many domains: sports tracking, physical rehabilitation, or real-time animation in entertainment.

Visual Question Answering automatically answers a question based on an image, combining computer vision and natural language processing. In some contexts, with targeted prompts, a Visual Question Answering model can extract more precise information than Image Captioning.

Our expertise for your project

We leverage the most effective pre-trained Deep Learning models on the market, such as ResNet classifiers, the YOLO detection model, or the SAM segmentation model. Depending on your product and business needs, we select the most suitable model and fine-tune it on your data to maximise performance.

Does your vision model need to be integrated into a business application? Accessible by hundreds or thousands of users while optimising hosting costs? Or downscaled for an embedded system with limited compute? Galadrim supports you at every stage of your project, with engineers experienced in integrating and deploying complex AI models.

Team

We are a team of 30 engineers and consultants passionate about artificial intelligence and the practical application of the latest research breakthroughs. Our mission: to democratise and simplify the use of AI for all our clients.

Let's discuss your project

Pablo, Partner & Head of AI

A graduate of HEC, Pablo leads Galadrim’s commercial development for AI and advises our clients on the strategic aspects of their projects. Before joining the tech industry, Pablo taught mathematics at higher education level.

View LinkedIn Profile

Benjamin, Partner & CTO AI

An engineer from École Polytechnique and the Corps des Mines, Benjamin oversees the technical team and advises our clients on the technologies best suited to their needs. After transitioning from senior civil service to entrepreneurship, he has held several Lead Technical roles.

View LinkedIn Profile

Our case studies

Image

Packaging model recognition

Development and fine-tuning of a model able to recognise one of Guillin’s 25,000 packaging references from an image.

Development of a dedicated mobile app for data collection and operational use of the model by the Group’s 250 sales representatives.
Adaptation and fine-tuning of pre-trained classification and detection models.
Deployment of the application and models on a suitable infrastructure.

Vidéo

Theft detection

Design and development of a real-time accident and theft detection system based on CCTV from autonomous stores.

End-to-end system design and oversight of suppliers in charge of hardware infrastructure (camera installation, etc.).
Adaptation and fine-tuning of pre-trained detection and pose estimation models.
Development of a web application enabling security staff to view real-time alerts and intervene remotely in stores.
Video stream orchestration and deployment of the solution on a suitable infrastructure.

Sémantique

Image search

In collaboration with YourArt, design and development of an AI feature integrated into the yourart.art platform, enabling users to find artworks from text or image queries.

Implementation of image and text semantic embedding models to generate vector representations.
Development of a nearest-neighbour search algorithm (ANN search).

What is Computer Vision?

Our expertise for your project

Team

Our case studies

Our technologies