Share

PIXANO simplifies image annotation for computer vision

#AI #computer vision #image annotation #machine learning #open source

PIXANO is an open-source solution for large-scale web annotation of pictures and videos for computer vision applications. PIXANO helps humans label images using semi-automated annotation.

Use

Smart annotation of visual data

Computer vision is one of the best-known applications of AI, enabling computers to automatically recognize objects in real and computer-generated images and in videos. Most perception methods currently in use require prior annotation of a large corpus of images used as examples during the learning phase.

PIXANO (Pixel Annotation) is an open-source solution facilitating large-scale web annotation of images and videos. It consists of a set of software tools and computer vision and machine learning algorithms that help humans optimize the process of creating labels and placing them on images.

Tool features:

  • Automated pre-annotation of images.
  • Assisted, interactive annotation, allowing users to make corrections.
  • Automatic temporal propagation for video annotation.
  • Active, incremental learning as new data, selected by an algorithm, is added.

 

PIXANO also includes powerful data analysis tools, enabling it to quickly search through a database of annotated images and select the data most relevant for any given task.

PIXANO has been used in projects with our R&D partners and in multi-partner projects like the EU H2020 CloudLSVA project. The solution has proven to be extremely effective in tests with major industrial companies, including in the automotive industry.

With its modular architecture and ability to integrate new, customizable smart components, PIXANO can help create a whole new range of solutions to meet AI developers’ needs.

Discover PIXANO

A powerful, open-source solution

Main advantages:

  • A wide range of label formats (bounding boxes, polygons, cuboids, landmarks, masks, etc.).
  • Data mining to build large corpora of annotated images for learning and evaluation enabled by the software’s advanced exploration criteria (based on semantic information or on classifications of visual content relevance based on a given vision task, for example).
  • Open-source and modular—PIXANO is based on a set of open-source, reusable, customizable, and scalable components.

Perception and data management solutionsPerception and data management solutions

  • Generating labeled databases for developing perception solutions for the manufacturing and other industries, as well as for security, robotics, transportation, etc.
  • Developing semi-automated annotation tools for labeling 2D or 3D images.
  • Developing a myriad of data management tools for data mining method exploration, dataset classification, computer-generated image production, and more.
Use cases

Autonomous vehicles

Large-scale automated annotation of data from cameras and LiDAR for autonomous vehicle perception systems.

Pixano | Smart annotation

Publications

UCP-Net: Unstructured Contour Points for Instance Segmentation, C. Dupont, Y. Ouakrim, Q. C. Pham, IEEE International Conference on Systems, Man, and Cybernetics (SMC), 2021.

Classifying All Interacting Pairs in a Single Shot, S. Chafik, A. Orcesi, R. Audigier and B. Luvison. IEEE Winter Conference on Applications of Computer Vision (WACV), 2020.

PandaNet: Anchor-Based Single-Shot Multi-Person 3D Pose Estimation, A. Benzine, F. Chabot, B. Luvison, Q. C. Pham, C. Achard, IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2020.

Deep MANTA: A Coarse-to-Fine Many-Task Network for Joint 2D and 3D Vehicle Analysis from Monocular Image, F. Chabot, M. Chaouch, J. Rabarisoa, C. Teulière and T. Chateau. IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), 2017.

Improving Multi-frame Data Association with Sparse Representations for Robust Near-online Multi-object Tracking, L. Fagot-Bouquet, R. Audigier, R. Dhome, F. Lerasle. European Conference on Computer Vision (ECCV), 2016.

Fast and accurate video annotation using dense motion hypotheses, L. Fagot-Bouquet, J. Rabarisoa, Q.C. Pham. IEEE International Conference on Image Processing (ICIP), 2014.

 

Visit PIXANO online

Use PIXANO

See also

#3D #AI #Blaxtair #image recognition
Success Story

Arcure
Arcure partnered with CEA-List to develop a vehicle-mounted 3D camera, used to prevent accidents by detecting pedestrians around industrial and construction vehicles.
Read more
#AI #data processing #frugality #trusted AI
Technology platforms

Artificial intelligence
Support the responsible development of AI-based systems for industry and society.
Read more