FocsleFo’c’sle
GO

EfficientDet-Lite4

4%
by Google

Reference EfficientDet variant pinned at INT8 for legacy NPUs. Maintained, not advanced.

Object detectionApache-2.0INT8cocolegacy
312K downloads 11K deploymentsUpdated Nov 8, 2027
Headline:19.2ms · Google Coral Edge TPU · INT8
OverviewBenchmarks6Sim ResultsDeploy6FilesDiscussion23

About this model

Reference EfficientDet variant pinned at INT8 for legacy NPUs. Maintained, not advanced.

Authored by google. Curated into the Fo’c’sle reference set on 2027-11-08. All cross-chip benchmarks below were collected in matched-pair runs in the HIL lab using the same input pipeline, same upstream preprocessing, and the same downstream consumer. See the methodology page for the full protocol.

Task
Object detection
Parameters
15.1 M
Benchmarked on
6 chips
Deployments
11K

Architecture

Detection backbone + neck + head
Inferred from upstream weights · simplified
ImageCSPDarknet53PANet neckCls headBox headObj headNMS · DFL

Headline benchmarks

πRaspberry Pi 5 + Hailo HATINT8
15.5ms p50
64 FPS98.4% acc7.0 W
QQualcomm QCS6490INT8
27.0ms p50
37 FPS98.5% acc8.3 W
AAmbarella CV5INT8
27.1ms p50
37 FPS97.5% acc3.9 W

Training data

Pretrained on the upstream maintainer’s released checkpoint. Edge-distillation pass uses 2.4M frames from the Fo’c’sle distillation corpus (consented public data + opt-in publisher contributions). Quantization-aware fine-tune uses 320K calibration samples drawn from the target task’s eval domain.

  • Pretraining corpus: upstream maintainer release
  • Distillation corpus: 2,400,000 frames
  • Calibration set: 320,000 samples (per task)
  • Eval set: standard benchmark + matched-pair HIL runs