From 85fa7a351f49122d83b7a0cdf5ddf3f4b8decc92 Mon Sep 17 00:00:00 2001 From: Vishal Khan Date: Fri, 11 Sep 2026 16:00:48 +0100 Subject: [PATCH] Add MIT licence and a notice separating code from data This repository had no LICENSE file, so default copyright applied and nobody could legally reuse the code regardless of it being public. Adds the standard MIT licence, copyright 2026 Vishal Khan. Because this project deliberately commits no HAM10000 images and fetches them at runtime instead, an MIT file on its own could be read as covering the dataset too. NOTICE.md states the split directly: MIT covers the code in this repository, and the HAM10000 corpus the fetch step downloads is CC BY-NC 4.0, non-commercial only, credited to Tschandl, Rosendahl and Kittler, Harvard Dataverse doi:10.7910/DVN/DBW86T. The README already cited the dataset licence and the paper in its data section but said nothing about the code's own terms. It gains two lines pointing at NOTICE.md. No existing figure or sentence there was changed. Co-Authored-By: Claude Opus 5 --- LICENSE | 21 +++++++++++++++++++++ NOTICE.md | 27 +++++++++++++++++++++++++++ README.md | 3 +++ 3 files changed, 51 insertions(+) create mode 100644 LICENSE create mode 100644 NOTICE.md diff --git a/LICENSE b/LICENSE new file mode 100644 index 0000000..ff7cdba --- /dev/null +++ b/LICENSE @@ -0,0 +1,21 @@ +MIT License + +Copyright (c) 2026 Vishal Khan + +Permission is hereby granted, free of charge, to any person obtaining a copy +of this software and associated documentation files (the "Software"), to deal +in the Software without restriction, including without limitation the rights +to use, copy, modify, merge, publish, distribute, sublicense, and/or sell +copies of the Software, and to permit persons to whom the Software is +furnished to do so, subject to the following conditions: + +The above copyright notice and this permission notice shall be included in all +copies or substantial portions of the Software. + +THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR +IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, +FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE +AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER +LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, +OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE +SOFTWARE. diff --git a/NOTICE.md b/NOTICE.md new file mode 100644 index 0000000..eafaf64 --- /dev/null +++ b/NOTICE.md @@ -0,0 +1,27 @@ +# Notice + +Two different licences apply here, and they do not cover the same things. + +## Code + +Everything in this repository is MIT licensed. See [LICENSE](LICENSE). +Copyright (c) 2026 Vishal Khan. + +## Data + +The HAM10000 dataset is **not** in this repository and is **not** MIT licensed. +No image is committed here. `uv run lesion-split fetch` downloads the corpus at +runtime into a gitignored directory on your own machine, under the dataset's own +terms rather than this repository's. + +HAM10000 is licensed **CC BY-NC 4.0**, which permits non-commercial use only. +Harvard Dataverse version 4.0, `doi:10.7910/DVN/DBW86T`. + +Tschandl, P., Rosendahl, C. and Kittler, H. The HAM10000 dataset, a large +collection of multi-source dermatoscopic images of common pigmented skin +lesions. *Scientific Data* 5, 180161 (2018). + +The MIT grant on the code does not extend to the data, and the non-commercial +restriction on the data does not restrict the code. If you run the training +arms you are using CC BY-NC 4.0 material and its terms apply to what you do +with it. diff --git a/README.md b/README.md index 86dbba1..662fac2 100644 --- a/README.md +++ b/README.md @@ -221,6 +221,9 @@ Nothing downloaded is committed. Non-commercial terms and a public repository do not mix, so the images land in a gitignored directory on the reader's machine under whatever terms they accepted. +The code in this repository is MIT licensed and the dataset is not. [NOTICE.md](NOTICE.md) +says which licence covers what. + Kaggle hosts this dataset in at least a dozen forms: balanced, augmented, hair removed, resized to 28 by 28. Training on a pre-augmented copy leaks by construction, because derived images of one original sit on both sides of any