[ Tech stack ]
Detectron2
Meta AI's PyTorch library for object detection and instance segmentation.
Detectron2 is Meta AI's object detection and segmentation library, written in PyTorch: its model zoo ships Faster R-CNN, Mask R-CNN, RetinaNet, Cascade R-CNN, Panoptic FPN and Keypoint R-CNN pretrained on COCO. A COCO-format dataset registers in the DatasetCatalog, a YAML or LazyConfig configuration drives training, and models export to TorchScript. It's our choice when the job calls for accurate instance masks or panoptic segmentation, rather than the raw speed of a real-time detector.
[ Why Detectron2 at Dexon ]
What this technology does well, and why we use it.
Typical usage: Industrial quality control, instance segmentation in medical imaging, object counting on aerial photos.
- 01
Model zoo: pretrained Mask R-CNN, Faster R-CNN, RetinaNet, Panoptic FPN.
- 02
Detection, instance segmentation, panoptic and keypoints in one codebase.
- 03
Training on custom COCO datasets, modular backbones and heads, built-in evaluation.
- 04
TorchScript export, Apache 2.0 license, foundation of Mask2Former and Detic.
[ Complementary technologies ]
The building blocks we often mobilise alongside.
A stack rarely exists alone. Here are the technologies Dexon most often pairs with this one, through pipeline habits, usage similarity or internal mastery. Click on a building block to see its scope.
[ Reassurance ]
- 0+
- custom projects delivered
- 30
- engineers, designers, project managers
- 80 %
- from top French schools
- 24 h
- average reply time
[ They trust us ]
More than 100 French and European companies trust us
[ Press ]
They talk about us.
Nationwide coverage by BFM Business, Le Figaro, Challenges, La Tribune and CNews. An outside reading of our work and our innovations.
