effdet

What it is and what it does

EfficientDet is a PyTorch port of Google's official TensorFlow object detection architecture, designed to balance speed and accuracy across multiple model scales. It provides pretrained weights for COCO detection and supports training on custom datasets (COCO, VOC, OpenImages) with a flexible configuration system. The core innovation is the BiFPN (Bidirectional Feature Pyramid Network), which you can customize via config—switching between depthwise separable and standard convolutions, adjusting activation functions, and choosing from any timm backbone that supports feature extraction.

The package includes training and validation scripts, focal loss with label smoothing, data loaders for multiple dataset formats, and evaluation tools based on TensorFlow Models. It supports PyTorch 1.6+ features like native DDP, SyncBN, and AMP, plus optional TorchScript compilation for speed. However, maintenance is dormant as of August 2024, so it is stable for inference and training on existing PyTorch versions but will not receive updates for new PyTorch releases or dependency changes.

Use it for:

• Train an object detector on a custom dataset using a pretrained EfficientDet backbone and fine-tuning scripts
• Deploy a lightweight COCO-pretrained detector for real-time inference on edge devices or servers
• Experiment with different BiFPN configurations and backbone architectures without modifying core model code
• Evaluate detection performance on VOC 2007/2012 or OpenImages datasets using built-in evaluators
• Convert a trained model to TorchScript for production inference with reduced memory and latency

Worth the install?

AI-flagged interpretation of the facts on this page — verify before relying

PyTorch implementation of EfficientDet object detection models with configurable BiFPN, multiple backbones from timm, and support for training on COCO and other datasets.

Yes, if you need a stable, well-documented PyTorch object detector for inference or training on current PyTorch versions. The pretrained weights are solid (mAP scores published), the codebase is mature, and there are no known vulnerabilities. No, if you require active maintenance, support for the latest PyTorch features, or expect bug fixes—the project is dormant and last released in May 2023. For new projects, consider whether a more actively maintained detection library fits your timeline.

Install

effdet on PyPI

pip install effdet

uv add effdet

poetry add effdet

Installing effdet

Before you install

Low install friction with a pure Python wheel. Maintenance is dormant—last release was May 2023 and no commits since August 2024—so expect no active bug fixes or dependency updates, though the codebase is stable enough for inference and training on existing PyTorch versions.

License in practice

Apache 2.0 permissive license allows commercial and private use with minimal restrictions; you may use, modify, and distribute the code freely as long as you include the license notice.

Quickstart

pip install effdet torch torchvision timm pycocotools omegaconf

import effdet
from effdet import get_detection_model

model = get_detection_model('tf_efficientdet_d0', pretrained=True)
model.eval()

Requires torch and torchvision; GPU recommended for inference speed but not required. Image dimensions must be divisible by 128 on each axis.

Verify before relying

• Whether pretrained weights cover all model variants listed in the description or only a subset
• Current compatibility with PyTorch 2.0+ given dormant maintenance status since 2023
• Whether all timm backbones with feature_only support work as drop-in replacements or require configuration tuning

Package facts

Evidence: effdet-0.4.1-py3-none-any.whl

Keywords: pytorch, pretrained, efficientdet, efficientnet, bifpn, object, detection

Tags

More Software Development packages