Yolo8 Indonesian Food Detection v1

--- license: mit base_model: - Ultralytics/YOLOv8 tags: - indonesian - food - detection - yolov8 - computer-vision - object-detection --- # 🍜 YOLOv8 Indonesian Food Detection v1 This model is a **YOLOv8-based object detection model** trained to detect various **Indonesian and international food items** from images. It is trained using a COCO-format dataset converted into YOLO format and fine-tuned using Ultralytics YOLOv8. --- ## 📌 Model Details - **Architecture:** YOLOv8 (Ultralytics) - **Task:** Object Detection - **Framework:** PyTorch / Ultralytics - **Base Model:** YOLOv8n - **Classes:** 12 food categories - **Image Size:** 640x640 - **Training Epochs:** 50 --- ## 🍱 Classes The model can detect the following food categories: - 0: Ayam Goreng - 1: Bakso - 2: Capcay - 3: Mie Goreng - 4: Nasi Goreng - 5: Pempek - 6: Rendang Sapi - 7: Sate - 8: Tahu Goreng - 9: Tempe Goreng - 10: Terong Balado - 11: Tumis Kangkung --- ## 🚀 Usage ### Install dependencies ```bash pip install ultralytics ``` ```python from ultralytics import YOLO model = YOLO("https://huggingface.co/wuriyanto/yolo8-indonesian-food-detection-v1/resolve/main/model.pt") results = model("food.jpg", conf=0.25) import matplotlib.pyplot as plt plt.imshow(results[0].plot()[:, :, ::-1]) plt.axis("off") plt.show() ``` ## 📊 Example Output The model detects food items and draws bounding boxes around them. Example inference result:    ## 📚 Dataset - Source: COCO-format Indonesian Food Dataset (multi-source) - Converted to YOLO format for training - Includes mixed Indonesian + international food categories - Split: Train / Validation / Test ## ⚠️ Limitations This model has several known limitations: ❌ Dataset is imbalanced across classes some classes have very few samples (e.g., Capcay) ❌ Detection may be inaccurate for: visually similar foods (e.g., Tahu vs Tempe) small objects in complex scenes ❌ Model does NOT perform classification beyond bounding box detection ❌ No calorie estimation or nutritional inference ❌ Performance may degrade in: low light conditions heavy occlusion non-food or abstract images ## 💡 Future Improvements (v2 roadmap) - Dataset balancing (increase low-sample classes like Capcay) - Remove non-food or irrelevant classes - Improve annotation consistency -...