EfficientDet

Category: model.detection.2d
Version: 2.0.0
UUID: 43a469452e2f762065eb3d08ca4d3f014a579ef609308edcad0825cf3686d029
Execution context: leip_af
Choice Priority: 1.0

Subcomponent Parameters

Name	Synonyms	Allowable Categories
Optimizer	optimization.optimizer	optimization.optimizer.generic
Scheduler	optimization.scheduler	optimization.scheduler.generic

Value Parameters

Name	Synonyms	Type	Values
Input Height	model.input.height	scalar	int, min: 1
Input Width	model.input.width	scalar	int, min: 1
Model Architecture	model.variation, model.architecture, model.backbone	choice	47 choices
Use Pretrained Backbone	model.use_pretrained, model.backbone.use_pretrained	scalar	bool
Include decoding in Graph	model.include_decoding, export.include_decoding	scalar	bool

Constraints

1. EfficientDet pre-trained backbones require imagenet normalization (data.normalization == 'imagenet' or module.pretrained==False)
2. EfficientDet models benefit from using TopK filtering in the post-processor. A good value is typically between 500-5000 (post_processor.k_threshold > 0)

This component fits into

Name	UUID	Synonyms
Full Recipe	4511dd...	model

Extra options

Model Architecture

efficientdet_d0
efficientdet_d1
efficientdet_d2
efficientdet_d3
efficientdet_d4
efficientdet_d5
efficientdetv2_dt
efficientdetv2_ds
resdet50
cspresdet50
cspresdext50
cspresdext50pan
cspdarkdet53
cspdarkdet53m
mixdet_m
mixdet_l
mobiledetv2_110d
mobiledetv2_120d
mobiledetv3_large
efficientdet_q0
efficientdet_q1
efficientdet_q2
efficientdet_w0
efficientdet_es
efficientdet_em
efficientdet_lite0
tf_efficientdet_d0
tf_efficientdet_d1
tf_efficientdet_d2
tf_efficientdet_d3
tf_efficientdet_d4
tf_efficientdet_d5
tf_efficientdet_d6
tf_efficientdet_d7
tf_efficientdet_d7x
tf_efficientdet_d0_ap
tf_efficientdet_d1_ap
tf_efficientdet_d2_ap
tf_efficientdet_d3_ap
tf_efficientdet_d4_ap
tf_efficientdet_d5_ap
tf_efficientdet_lite0
tf_efficientdet_lite1
tf_efficientdet_lite2
tf_efficientdet_lite3
tf_efficientdet_lite3x
tf_efficientdet_lite4