added EHTHIC.md and MODEL_CARD.md

Author: codinglabsong

ETHICS.md

@@ -0,0 +1,23 @@
+# Ethical Considerations
+
+This project trains generative adversarial networks to translate face images between "young" and "old" domains. The code is research oriented and makes use of the publicly available **UTKFace** dataset. The following points summarize the main ethical issues and recommended practices.
+
+## Data Source
+* **UTKFace** images contain photos of real people. The dataset is distributed for academic purposes. Users must follow the dataset's license and terms of use.
+* Data were center cropped and resized to 256×256 pixels. Only age labels were used to create two groups: ages 18‑28 and ages 40 and above.
+
+## Intended Use
+* The software is meant for research and educational exploration of image‑to‑image translation techniques.
+* It should **not** be used for deception, surveillance, or impersonation. Generated images may not accurately reflect the depicted person.
+
+## Bias and Limitations
+* UTKFace is not perfectly balanced across ethnicities or genders. Models trained on it may reflect these imbalances.
+* The approach does not guarantee realistic aging effects and may fail on extreme poses or occlusions.
+
+## Responsible Usage
+* Do not use the model to create misleading or defamatory content.
+* Clearly disclose synthetic images in any public communication.
+* Comply with privacy regulations and obtain consent before processing personal images.
+
+## Contact
+Questions or concerns can be directed to the project maintainers through the repository's issue tracker.

MODEL_CARD.md

@@ -0,0 +1,25 @@
+# Model Card: Aging GAN
+
+Aging GAN is a CycleGAN‑style model that translates face images between two age domains: **Young** (ages 18‑28) and **Old** (ages 40+). It consists of two U‑Net generators and two PatchGAN discriminators trained with adversarial, cycle‑consistency and identity losses.
+
+## Intended Use
+* Research and educational experiments on image‑to‑image translation and face aging.
+* Not intended for production applications, facial recognition, or any form of impersonation or deceptive media generation.
+
+## Dataset
+* Training data is derived from the **UTKFace** dataset. Images were cropped and resized to 256×256 pixels.
+* The dataset was split 80/10/10 into train/validation/test subsets with equal numbers of young and old images.
+
+## Training
+* Default training runs for 100 epochs using the Adam optimizer and mixed‑precision via Hugging Face Accelerate.
+* Frechet Inception Distance (FID) is computed on the validation set each epoch; the model with the best FID score is saved.
+
+## Limitations
+* The model is limited to the diversity and quality of UTKFace and may not generalize to unseen demographics or lighting conditions.
+* Generated aging effects are approximate and should not be considered authentic depictions of an individual.
+
+## Ethical Considerations
+See [ETHICS.md](ETHICS.md) for discussion of data usage, potential biases and recommended precautions.
+
+## Citation
+If you use this code or model in your research, please cite the repository and the UTKFace dataset.