Unet Damaged Flower Segmentation

--- license: mit tags: - image-segmentation - unet - tensorflow - flower - damage-detection datasets: - metavision/accurate-damaged-flower-shapessegmentation --- # U-Net - Damaged Flower Segmentation Model U-Net untuk segmentasi area kerusakan pada bunga. ## Dataset Accurate Damaged Flower Shapes Segmentation (Kaggle) ## Arsitektur - Model: U-Net dengan 4 level encoder-decoder - Input: 256×256×3 (RGB) - Output: 256×256×1 (Binary mask) - Loss: BCE + Dice Loss ## Hasil Evaluasi | Metrik | Score | |-----------|--------| | Accuracy | 0.9998 | | F1 Score | 0.0000 | | IoU | 0.0000 | | Dice | 0.0000 | ## Penggunaan ```python from tensorflow.keras.models import load_model import numpy as np from PIL import Image model = load_model('unet_flower_segmentation.h5', ...) img = np.array(Image.open('bunga.jpg').resize((256, 256))) / 255.0 mask = model.predict(img[np.newaxis, ...]) ```