🎨 Anime Autoencoder Playground – by a 16-Year-Old Enthusiast

---

🌟 Introduction

Welcome! This project is a fun, experimental AI project created by a 16-year-old enthusiast.
The goal is to explore image generation using deep autoencoders trained on anime images.

⚠️ Disclaimer:
This project is purely for learning and experimentation.
Models are simple, training is short, and outputs are illustrative.
Feel free to customize, extend, and experiment – this is meant as a playground. 😎
The results are not production-grade and mainly serve as a guideline/example.

---

📂 File Structure

├── DataF/ # Downloaded anime images (~30 images) ├── Model/ # Saved autoencoder model ├── datapng/ # Generated sample images ├── main.py # Main training and generation script └── README.md # This README file

markdown คัดลอกโค้ด

---

⚙️ How It Works – Step by Step

1️⃣ Fetch Images

• Uses Google Custom Search API
• Queries: "anime character", "anime background", "anime scenery"
• Downloads ~30 images into DataF/

2️⃣ Data Preprocessing

• Resize images to 512×512 pixels
• Normalize pixel values to [0,1]
• Automatically skip corrupted files

3️⃣ Build Autoencoder

• Encoder: 3× Conv2D + MaxPooling layers
• Decoder: 3× Conv2D + UpSampling layers
• Output: reconstructed RGB image

4️⃣ Training

• Loss: Mean Squared Error (MSE)
• Optimizer: Adam
• Batch size: 1
• Shuffle dataset every epoch
• Trains for 20 epochs (simple/fun experiment)

5️⃣ Save Model

• Model saved to Model/autoencoder_anime_512.h5

6️⃣ Generate Image

• Decode a batch from trained autoencoder
• Save output to datapng/generated_anime_512.png

---

🛠️ Customization & Tips

• Change queries to fetch different types of images
• Increase dataset size for better quality
• Modify encoder/decoder layers for deeper/wider architecture
• Experiment with data augmentation or other generative models (GANs, VQ-VAE)
• Feel free to fork and adjust this project to your liking

---

💖 Support Me on GitHub Sponsors

If you enjoy this project, support me here:

Or click directly: https://github.com/sponsors/FWKMultiverse

Your support helps me upgrade hardware, run AI experiments, and continue building projects. Every contribution means a lot! 🚀

---

📬 Contact

• 📧 Email: yoglawm644@gmail.com
• 🌐 Facebook Page: FWK Multiverse
• 🐦 Twitter/X: @FWK_Multiverse

---

🏷️ Tags & Topics

AI Deep Learning Autoencoder Anime Image Generation Python TensorFlow Experiment Playground Beginner-Friendly

---

⚡ Summary

• Purpose: Learning AI image generation through a fun playground project
• Not production-grade: Intended as a simple example for learning
• Codebase: Easy to adapt, extend, and experiment with
• Audience: Beginners, hobbyists, students, or anyone curious about AI

🎉 Enjoy exploring AI, and feel free to fork, modify, and experiment!
Remember: this is a learning playground, results are simple but fun. 😎