README.md

metadata

title: Mülltrenner9000
emoji: ♻️
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 5.34.2
app_file: app.py
pinned: false

Mülltrenner9000 ♻️

Bin there, sorted that.

An instance segmentation app for waste classification, built with Gradio and YOLOv11.

Mülltrenner9000 is a smart instance segmentation web app that classifies and segments waste into 59 categories, recommending the correct German recycling bin—such as Gelbe Tonne, Restmüll, or Pfand—based on real-time object detection.

The model uses a fine-tuned YOLOv11 segmentation model (YOLO11m-seg) and offers an interactive Gradio-based UI for waste classification and bin assignment.

---

🔗 Try it on Hugging Face Spaces: Mülltrenner9000 – Hugging Face Space

---

🧠 What It Does

• ✅ Instance Segmentation of individual waste items
• ✅ Classification into correct bins:
  • Yellow Bin (Gelbe Tonne) – Plastics & Metals
  • Grey Bin (Restmüll) – General Waste
  • Green Bin (Biotonne) – Biodegradable Waste
  • Blue Bin (Papiertonne) – Paper & Cardboard
  • Glascontainer – Glass Waste
  • Hazardous Waste (Sondermüll) – e.g. Batteries, Chemicals
  • Deposit Return (Pfand) – Refundable items
• ✅ Real-time segmentation masks with bin color overlays and labels

---

🗂 Dataset

• Source: TACO: Trash Annotations in Context (Roboflow)
• Annotations: Pixel-level masks + class labels
• Images: 3147 for training, 299 for validation
• Classes: 59 object types

---

🛠 Model Training

• Model: YOLOv11 (YOLO11m-seg)
• Backbone: Fine-tuned with heavy data augmentation (RandAugment, CLAHE, etc.)
• Epochs: Trained for 70 epochs
• Batch Size: 16
• Optimizer: SGD
• Learning Rate: Final stage LR tuned to 0.001
• Mixed Precision: AMP enabled

---

🖼 Sample Output

Here is a sample segmentation and classification output of Mülltrenner9000:

---

🤝 Contribution

We welcome contributions from the community, whether it's in the form of bug reports, improvements, documentation, or new features.

---

📄 License

This project is licensed under the MIT License. See the LICENSE file for more details.