Waste Classification V2 (EfficientNet)
30-class waste image classifier built with EfficientNet transfer learning.
Model Details
- Architecture: EfficientNetB0 + custom classification head
- Input size: 224 x 224 x 3
- Framework: TensorFlow / Keras
- Author: Darshan764
Performance
- Validation Accuracy: 87.3%
- Validation set: 3,000 images (20% held-out)
- Number of classes: 30
Classes
aerosol_cans, aluminum_food_cans, aluminum_soda_cans, cardboard_boxes, cardboard_packaging, clothing, coffee_grounds, disposable_plastic_cutlery, eggshells, food_waste, glass_beverage_bottles, glass_cosmetic_containers, glass_food_jars, magazines, newspaper, office_paper, paper_cups, plastic_cup_lids, plastic_detergent_bottles, plastic_food_containers, plastic_shopping_bags, plastic_soda_bottles, plastic_straws, plastic_trash_bags, plastic_water_bottles, shoes, steel_food_cans, styrofoam_cups, styrofoam_food_containers, tea_bags
Dataset
Recyclable and Household Waste Classification
- 15,000 images total (250 per class)
- 80/20 train/validation split (seed=42)
Training
- Optimizer: Adam
- Loss: sparse_categorical_crossentropy
- Batch size: 32
- Callbacks: EarlyStopping, ReduceLROnPlateau, ModelCheckpoint
How to Use
import tensorflow as tf
import numpy as np
from tensorflow.keras.utils import load_img, img_to_array
model = tf.keras.models.load_model("hf://Darshan764/waste-classification-v2")
img = load_img("waste.jpg", target_size=(224, 224))
arr = np.expand_dims(img_to_array(img), axis=0)
pred = model.predict(arr, verbose=0)
idx = int(np.argmax(pred[0]))
print(f"Class index: {idx}")
print(f"Confidence: {pred[0][idx]*100:.2f}%")
Limitations
- Trained only on the Kaggle dataset above; may not generalize to very different lighting/backgrounds.
- Not intended for production waste-sorting systems without further validation.
language: en license: apache-2.0 library_name: keras tags: - image-classification - waste-classification - tensorflow - keras - efficientnet datasets: - alistairking/recyclable-and-household-waste-classification metrics: - accuracy
Waste Classification V2 (EfficientNet)
30-class waste image classifier built with EfficientNet transfer learning.
Model Details
- Architecture: EfficientNetB0 + custom classification head
- Input size: 224 x 224 x 3
- Framework: TensorFlow / Keras
- Author: Darshan764
Performance
- Validation Accuracy: 87.3%
- Validation set: 3,000 images (20% held-out)
- Number of classes: 30
Classes
aerosol_cans, aluminum_food_cans, aluminum_soda_cans, cardboard_boxes, cardboard_packaging, clothing, coffee_grounds, disposable_plastic_cutlery, eggshells, food_waste, glass_beverage_bottles, glass_cosmetic_containers, glass_food_jars, magazines, newspaper, office_paper, paper_cups, plastic_cup_lids, plastic_detergent_bottles, plastic_food_containers, plastic_shopping_bags, plastic_soda_bottles, plastic_straws, plastic_trash_bags, plastic_water_bottles, shoes, steel_food_cans, styrofoam_cups, styrofoam_food_containers, tea_bags
Dataset
Recyclable and Household Waste Classification
- 15,000 images total (250 per class)
- 80/20 train/validation split (seed=42)
Training
- Optimizer: Adam
- Loss: sparse_categorical_crossentropy
- Batch size: 32
- Callbacks: EarlyStopping, ReduceLROnPlateau, ModelCheckpoint
How to Use
import tensorflow as tf
import numpy as np
from tensorflow.keras.utils import load_img, img_to_array
model = tf.keras.models.load_model("hf://Darshan764/waste-classification-v2")
img = load_img("waste.jpg", target_size=(224, 224))
arr = np.expand_dims(img_to_array(img), axis=0)
pred = model.predict(arr, verbose=0)
idx = int(np.argmax(pred[0]))
print(f"Class index: {idx}")
print(f"Confidence: {pred[0][idx]*100:.2f}%")
Limitations
- Trained only on the Kaggle dataset above; may not generalize to very different lighting/backgrounds.
- Not intended for production waste-sorting systems without further validation.
🎮 Live Demo
Try the model in your browser (upload any waste image):
Sample verification — uploaded a real-world plastic water bottle image:
| Rank | Class | Confidence |
|---|---|---|
| 1 | plastic_water_bottles | 92.7% |
| 2 | plastic_soda_bottles | 6.9% |
| 3 | plastic_trash_bags | 0.2% |
- Downloads last month
- 31