DETR

GPTKB entity

Statements (42)
Predicate Object
gptkbp:instanceOf gptkb:model
gptkbp:architecture gptkb:transformation
gptkbp:author gptkb:Nicolas_Usunier
gptkb:Francisco_Massa
gptkb:Alexander_Kirillov
gptkb:Gabriel_Synnaeve
gptkb:Nicolas_Carion
gptkb:Sergey_Zagoruyko
gptkbp:category gptkb:artificial_intelligence
gptkb:machine_learning
computer vision
gptkbp:citation over 3000 (as of 2023)
gptkbp:developedBy gptkb:Facebook_AI_Research
gptkbp:features end-to-end object detection
no need for anchor boxes
no need for non-maximum suppression
gptkbp:fullName gptkb:Detection_Transformer
gptkbp:github https://github.com/facebookresearch/detr
https://www.w3.org/2000/01/rdf-schema#label DETR
gptkbp:input gptkb:illustrator
gptkbp:inspiredBy transformer architecture
gptkbp:introducedIn 2020
gptkbp:language gptkb:Python
gptkbp:license gptkb:MIT_License
gptkbp:limitation slow convergence
difficulty with small objects
gptkbp:notableContributor first to use transformers for object detection
gptkbp:notablePublication gptkb:End-to-End_Object_Detection_with_Transformers
gptkbp:openSource gptkb:PyTorch
gptkbp:output class labels
bounding boxes
gptkbp:performanceContext competitive with Faster R-CNN
gptkbp:publicationDate gptkb:European_Conference_on_Computer_Vision_(ECCV)_2020
gptkbp:relatedModel gptkb:Conditional_DETR
gptkb:DN-DETR
gptkb:Deformable_DETR
gptkbp:trainer gptkb:COCO
gptkbp:trainingTime longer than traditional detectors
gptkbp:usedFor image segmentation
object detection
gptkbp:bfsParent gptkb:Department_for_the_Environment,_Transport_and_the_Regions
gptkbp:bfsLayer 5