LibreEfficientDet third-party attribution
=========================================

The EfficientDet D0-D4 architecture, TensorFlow-compatible EfficientNet
backbones, BiFPN topology, prediction heads, anchor generation, checkpoint
configuration, and evaluation transform are derived from:

    Repository: https://github.com/rwightman/efficientdet-pytorch
    Commit:     c6dff775a36cea0bf9b76c58e59f936411c5ce01
    Release:    effdet 0.4.1
    License:    Apache License 2.0
    Copyright:  Copyright 2020 Ross Wightman

The native EfficientNet block structure and timm-compatible parameter names
also follow:

    Repository: https://github.com/huggingface/pytorch-image-models
    Version:    v1.0.28
    Commit:     8ef73809f622e0031bd7f4940265734aef8b9978
    License:    Apache License 2.0
    Copyright:  Ross Wightman and the timm contributors

The original EfficientDet TensorFlow implementation is from Google Research:

    Repository: https://github.com/google/automl/tree/master/efficientdet
    BiFPN source commit: 56815c9986ffd4b508fe1d68508e268d129715c1
    Anchor source commit: 6f6694cec1a48cdb33d5d1551a2d5db8ad227798
    License: Apache License 2.0
    Copyright: Copyright 2020 Google Research

Official D0-D4 COCO checkpoints are unmodified learned tensors from the
``rwightman/efficientdet-pytorch`` v0.1 GitHub release. That Apache-2.0
project supplies no checkpoint-specific license, so the redistribution basis
is Apache-2.0 implied by the releasing project rather than a separate explicit
weight grant. Converted files add LibreYOLO metadata only.

No source from the LGPL-licensed
``zylo117/Yet-Another-EfficientDet-Pytorch`` project was consulted or used.

The Apache License 2.0 text is distributed at
``licenses/Apache-2.0.txt``.