Instructions to use ashuwhy/efficientnetv2s with libraries, inference providers, notebooks, and local apps. Follow these links to get started.
- Libraries
- Keras
How to use ashuwhy/efficientnetv2s with Keras:
# !pip install -U keras tensorflow huggingface_hub # Keras needs TensorFlow installed to read "hf://" paths, so the tensorflow backend is selected here; # "jax" and "torch" also work for computation once TensorFlow is installed. import os os.environ["KERAS_BACKEND"] = "tensorflow" import keras model = keras.saving.load_model("hf://ashuwhy/efficientnetv2s") - Notebooks
- Google Colab
- Kaggle
Upload 2 files
Browse files- convert_model.py +94 -0
- download_models.py +66 -0
convert_model.py
ADDED
|
@@ -0,0 +1,94 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Convert script for waste classification models.
|
| 4 |
+
This script converts the model to a compatible format for TensorFlow 2.12.0.
|
| 5 |
+
"""
|
| 6 |
+
|
| 7 |
+
import os
|
| 8 |
+
import sys
|
| 9 |
+
import tensorflow as tf
|
| 10 |
+
import numpy as np
|
| 11 |
+
from tensorflow.keras.applications import EfficientNetV2S
|
| 12 |
+
from tensorflow.keras.layers import Dense, GlobalAveragePooling2D, Input
|
| 13 |
+
from tensorflow.keras.models import Model
|
| 14 |
+
|
| 15 |
+
def create_compatible_model(num_classes=7):
|
| 16 |
+
"""Create a compatible model with the same architecture"""
|
| 17 |
+
# Create input layer explicitly
|
| 18 |
+
input_tensor = Input(shape=(224, 224, 3))
|
| 19 |
+
|
| 20 |
+
# Use EfficientNetV2S as base model
|
| 21 |
+
base_model = EfficientNetV2S(
|
| 22 |
+
include_top=False,
|
| 23 |
+
weights='imagenet',
|
| 24 |
+
input_tensor=input_tensor
|
| 25 |
+
)
|
| 26 |
+
|
| 27 |
+
# Freeze the base model layers
|
| 28 |
+
for layer in base_model.layers:
|
| 29 |
+
layer.trainable = False
|
| 30 |
+
|
| 31 |
+
# Add custom classification head
|
| 32 |
+
x = base_model.output
|
| 33 |
+
x = GlobalAveragePooling2D()(x)
|
| 34 |
+
x = Dense(512, activation='relu')(x)
|
| 35 |
+
predictions = Dense(num_classes, activation='softmax')(x)
|
| 36 |
+
|
| 37 |
+
model = Model(inputs=input_tensor, outputs=predictions)
|
| 38 |
+
|
| 39 |
+
# Compile the model
|
| 40 |
+
model.compile(
|
| 41 |
+
optimizer='adam',
|
| 42 |
+
loss='categorical_crossentropy',
|
| 43 |
+
metrics=['accuracy']
|
| 44 |
+
)
|
| 45 |
+
|
| 46 |
+
return model
|
| 47 |
+
|
| 48 |
+
def convert_model(source_path, target_path):
|
| 49 |
+
"""Convert model to a compatible format"""
|
| 50 |
+
print(f"Creating a compatible model...")
|
| 51 |
+
model = create_compatible_model()
|
| 52 |
+
|
| 53 |
+
print(f"Saving compatible model to {target_path}...")
|
| 54 |
+
model.save(target_path)
|
| 55 |
+
|
| 56 |
+
print(f"Model saved successfully!")
|
| 57 |
+
return True
|
| 58 |
+
|
| 59 |
+
def main():
|
| 60 |
+
import argparse
|
| 61 |
+
parser = argparse.ArgumentParser(description='Convert model to a compatible format')
|
| 62 |
+
parser.add_argument('--source', type=str, default='efficientnetv2s_waste_classifier_final.keras',
|
| 63 |
+
help='Source model path')
|
| 64 |
+
parser.add_argument('--target', type=str, default='efficientnetv2s_waste_classifier_compatible.keras',
|
| 65 |
+
help='Target model path')
|
| 66 |
+
|
| 67 |
+
args = parser.parse_args()
|
| 68 |
+
|
| 69 |
+
# Get the directory of this script
|
| 70 |
+
script_dir = os.path.dirname(os.path.abspath(__file__))
|
| 71 |
+
|
| 72 |
+
# Resolve paths
|
| 73 |
+
source_path = os.path.join(script_dir, args.source)
|
| 74 |
+
target_path = os.path.join(script_dir, args.target)
|
| 75 |
+
|
| 76 |
+
# Check if source model exists
|
| 77 |
+
if not os.path.exists(source_path):
|
| 78 |
+
print(f"Source model not found at {source_path}")
|
| 79 |
+
return 1
|
| 80 |
+
|
| 81 |
+
# Convert model
|
| 82 |
+
success = convert_model(source_path, target_path)
|
| 83 |
+
|
| 84 |
+
if success:
|
| 85 |
+
print(f"Model converted successfully!")
|
| 86 |
+
print(f"Source: {source_path}")
|
| 87 |
+
print(f"Target: {target_path}")
|
| 88 |
+
return 0
|
| 89 |
+
else:
|
| 90 |
+
print(f"Model conversion failed!")
|
| 91 |
+
return 1
|
| 92 |
+
|
| 93 |
+
if __name__ == "__main__":
|
| 94 |
+
sys.exit(main())
|
download_models.py
ADDED
|
@@ -0,0 +1,66 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 1 |
+
#!/usr/bin/env python3
|
| 2 |
+
"""
|
| 3 |
+
Download script for waste classification models.
|
| 4 |
+
This script downloads the model files from HuggingFace if they don't exist locally.
|
| 5 |
+
"""
|
| 6 |
+
|
| 7 |
+
import os
|
| 8 |
+
import sys
|
| 9 |
+
import requests
|
| 10 |
+
from tqdm import tqdm
|
| 11 |
+
|
| 12 |
+
# Model URLs from HuggingFace
|
| 13 |
+
MODEL_URLS = {
|
| 14 |
+
"efficientnetv2s_waste_classifier.keras": "https://huggingface.co/ashuwhy/efficientnetv2s/resolve/main/efficientnetv2s_waste_classifier.keras",
|
| 15 |
+
"efficientnetv2s_waste_classifier_fine_tuned.keras": "https://huggingface.co/ashuwhy/efficientnetv2s/resolve/main/efficientnetv2s_waste_classifier_fine_tuned.keras",
|
| 16 |
+
"efficientnetv2s_waste_classifier_final.keras": "https://huggingface.co/ashuwhy/efficientnetv2s/resolve/main/efficientnetv2s_waste_classifier_final.keras"
|
| 17 |
+
}
|
| 18 |
+
|
| 19 |
+
def download_file(url, destination):
|
| 20 |
+
"""Download a file with progress bar"""
|
| 21 |
+
response = requests.get(url, stream=True)
|
| 22 |
+
total_size = int(response.headers.get('content-length', 0))
|
| 23 |
+
block_size = 1024 # 1 Kibibyte
|
| 24 |
+
|
| 25 |
+
print(f"Downloading {os.path.basename(destination)} ({total_size / (1024*1024):.2f} MB)")
|
| 26 |
+
|
| 27 |
+
with open(destination, 'wb') as file, tqdm(
|
| 28 |
+
desc=os.path.basename(destination),
|
| 29 |
+
total=total_size,
|
| 30 |
+
unit='iB',
|
| 31 |
+
unit_scale=True,
|
| 32 |
+
unit_divisor=1024,
|
| 33 |
+
) as bar:
|
| 34 |
+
for data in response.iter_content(block_size):
|
| 35 |
+
size = file.write(data)
|
| 36 |
+
bar.update(size)
|
| 37 |
+
|
| 38 |
+
print(f"Downloaded {os.path.basename(destination)} to {destination}")
|
| 39 |
+
|
| 40 |
+
def main():
|
| 41 |
+
# Get the directory of this script
|
| 42 |
+
script_dir = os.path.dirname(os.path.abspath(__file__))
|
| 43 |
+
|
| 44 |
+
# Check and download each model
|
| 45 |
+
for model_name, model_url in MODEL_URLS.items():
|
| 46 |
+
model_path = os.path.join(script_dir, model_name)
|
| 47 |
+
|
| 48 |
+
if os.path.exists(model_path):
|
| 49 |
+
print(f"Model {model_name} already exists at {model_path}")
|
| 50 |
+
else:
|
| 51 |
+
print(f"Model {model_name} not found. Downloading...")
|
| 52 |
+
try:
|
| 53 |
+
download_file(model_url, model_path)
|
| 54 |
+
except Exception as e:
|
| 55 |
+
print(f"Error downloading {model_name}: {e}")
|
| 56 |
+
|
| 57 |
+
print("\nDownload complete. Available models:")
|
| 58 |
+
for model_name in MODEL_URLS.keys():
|
| 59 |
+
model_path = os.path.join(script_dir, model_name)
|
| 60 |
+
if os.path.exists(model_path):
|
| 61 |
+
print(f"✅ {model_name} ({os.path.getsize(model_path) / (1024*1024):.2f} MB)")
|
| 62 |
+
else:
|
| 63 |
+
print(f"❌ {model_name} (not available)")
|
| 64 |
+
|
| 65 |
+
if __name__ == "__main__":
|
| 66 |
+
main()
|