THIRD-PARTY RIGHTS AND DATA USE NOTICE
Date: 2026-08-04

Project scope
-------------
This project is an engineering and research demonstration for optical screening of
rice areas whose imagery may be consistent with bacterial leaf blight (BLB). It is
not a pathogen diagnosis, food-safety determination, pesticide prescription, or
autonomous-treatment system.

Project-created material
------------------------
Version 1.0 (DOI 10.5281/zenodo.21787704) applies CC BY 4.0 only to the
author-owned original analysis, text, research software, and original figures.
That license does not relicense third-party datasets, articles, documentation,
trademarks, URLs, images, text, software, or other materials.

Packaged font
-------------
The packaged Noto Sans TC font and its OFL.txt license are third-party materials.
The font remains governed by the SIL Open Font License and is not relicensed under
this project's CC BY 4.0 notice.

BLB UAV Dataset
---------------
The Figshare record "Bacterial Leaf Blight (BLB) UAV Dataset and U-Net with
ResNet-101 Code for BLB Detection", version 3, is identified by:

    https://doi.org/10.6084/m9.figshare.26955862

The repository record displays Creative Commons Attribution 4.0 International.
Reuse must retain creator attribution, the DOI, the license notice, and an
indication of project changes. The dataset was captured with DJI Phantom 4
Multispectral, not DJI Mavic 3M. Its original split is used only as a reproduction
benchmark; it is not represented as Taiwan or cross-field validation.

GD-RGBMS-DB
-----------
The Zenodo record "GD-RGBMS-DB: UAV RGB and Multispectral Dataset for Rice Grain
Discoloration Segmentation" is identified by:

    https://doi.org/10.5281/zenodo.17008818

Its License field was blank when checked on 2026-08-04. Open access to a record
does not by itself grant permission to copy, redistribute, train a model, publish
derived weights, or use the material commercially. This source is therefore a
Mavic 3M pipeline-compatibility candidate only. No files, previews, training
outputs, or weights derived from it may be publicly released until explicit
rights are documented. This notice does not describe the record as CC BY 4.0.

NCHU rice nematode UAV dataset
------------------------------
The National Chung Hsing University dataset page is:

    https://aidata.nchu.edu.tw/smarter/dataset/smarter_06_nema_0_rgb_20250606_1234_bessyi

The record-specific resource page states that no usable license is available, and
the platform conditions restrict use and redistribution. No permission is assumed
for this project. The data must not be placed in Git, a website, a public research
package, a commercial workflow, or public model weights without written approval
covering each intended use.

Other third-party sources
-------------------------
Article licenses and dataset licenses are evaluated separately. An open article
does not make its underlying data open. Data described as "available on request"
are not included without permission. DJI manuals and webpages are cited and
linked as official operational references; they are not redistributed and do not
constitute independent validation of disease-detection accuracy. DJI and Mavic are
third-party names and trademarks used only to identify equipment. This project is
not affiliated with, sponsored by, or endorsed by DJI, Figshare, Zenodo, NCHU,
PLOS, MDPI, Mendeley, or any named data creator.

Models and weights
------------------
A model file may reflect its training inputs. Model weights will be released only
when every source in the exact training manifest permits the intended training,
derivative publication, redistribution, and use. Rights-unresolved or restricted
data are excluded from release-bound training. Publication of weights does not
transfer third-party rights in training data or authorize diagnostic, regulatory,
food-safety, or treatment claims.

Privacy and locations
---------------------
Public packages exclude personal information and precise private-field locations
unless informed, written permission expressly authorizes disclosure. Raw captures
that contain people, vehicle plates, neighboring property, or identifiable
landholder information are restricted, cropped, masked, or withheld.

Corrections and requests
------------------------
A rights holder or data subject who identifies an error, missing attribution, or
unauthorized inclusion may request review through the contact method published
with the research record. Verified issues will be handled through a versioned
correction, restriction, replacement, or removal process. This notice reduces
ambiguity; it does not claim to eliminate all legal or research risk.
