Skip to content

D-FINE

Category: model.detection.2d
Version: 1.0.0
UUID: 06ff3d46b7a3a84c1d859c3366b6d326ae967047aa37deed7e72780681f24f44
Execution context: leip_af
Choice Priority: 0.5

Subcomponent Parameters

Name Synonyms Allowable Categories
Optimizer optimization.optimizer optimization.optimizer.rtdetr

Value Parameters

Name Synonyms Type Values
Input Height model.input.height scalar int, min: 1
Input Width model.input.width scalar int, min: 1
Freeze Backbone model.backbone.freeze scalar bool
Learning rate optimization.scheduler.final_learning_rate scalar float
Model Architecture model.variation,
model.backbone,
model.architecture
choice 9 choices
Top queries to use post_processor.max_detections,
post_processor.max_det_per_image
scalar int, min: 0
Use Pretrained Backbone model.use_pretrained,
model.backbone.use_pretrained
scalar bool

Constraints

  1. Data for D-FINE models should be normalized with 'yolo' (data.normalization == 'yolo')

This component fits into

Name UUID Synonyms
Full Recipe 8b4b05... model

Extra options

Model Architecture

D-FINE-N (Nano)
D-FINE-S (Small)
D-FINE-M (Medium)
D-FINE-L (Large)
D-FINE-X (Extra Large)
D-FINE-S-O (Small - Objects365)
D-FINE-M-O (Medium - Objects365)
D-FINE-L-O (Large - Objects365)
D-FINE-X-O (Extra Large - Objects365)