CGI_Classification / cgi_classification_app.py
nhradek's picture
Upload folder using huggingface_hub
16ff589 verified
raw
history blame
3.87 kB
# -*- coding: utf-8 -*-
"""CGI Classification App.ipynb
Automatically generated by Colab.
Original file is located at
https://colab.research.google.com/drive/1ckzOtXUiFW_NqlIandwoH07lnsLGKTLB
"""
from scipy.spatial import distance
import numpy as np
class MeanClassifier:
def fit(self, X, y):
self.mean_0 = np.mean(X[y == 0], axis=0) if np.any(y == 0) else None
self.mean_1 = np.mean(X[y == 1], axis=0) if np.any(y == 1) else None
def predict(self, X):
preds = []
for x in X:
dist_0 = (
distance.euclidean(x, self.mean_0)
if self.mean_0 is not None
else np.inf
)
dist_1 = (
distance.euclidean(x, self.mean_1)
if self.mean_1 is not None
else np.inf
)
preds.append(1 if dist_1 < dist_0 else 0)
return np.array(preds)
def predict_proba(self, X):
# An implementation of probability prediction which uses a softmax function to determine the probability of each class based on the distance to the mean for each prototype
preds = []
for x in X:
dist_0 = (
distance.euclidean(x, self.mean_0) if self.mean_0 is not None else np
)
dist_1 = (
distance.euclidean(x, self.mean_1)
if self.mean_1 is not None
else np.inf
)
prob_0 = np.exp(-dist_0) / (np.exp(-dist_0) + np.exp(-dist_1))
prob_1 = np.exp(-dist_1) / (np.exp(-dist_0) + np.exp(-dist_1))
preds.append([prob_0, prob_1])
return np.array(preds)
def mean_distance(self, x):
dist_mean_0 = (
distance.euclidean(x, self.mean_0) if self.mean_0 is not None else np.inf
)
dist_mean_1 = (
distance.euclidean(x, self.mean_1) if self.mean_1 is not None else np.inf
)
return dist_mean_0, dist_mean_1
import gradio as gr
from PIL import Image
import numpy as np
from PIL import Image
from scipy.fftpack import fft2
from tensorflow.keras.models import load_model, Model
import pickle
mean_clf = None
with open("mean_clf.pkl", "rb") as f:
mean_clf = pickle.load(f)
# Function to apply Fourier transform
def apply_fourier_transform(image):
image = np.array(image)
fft_image = fft2(image)
return np.abs(fft_image)
def preprocess_image(image):
try:
image = Image.fromarray(image)
image = image.convert("L")
image = image.resize((256, 256))
image = apply_fourier_transform(image)
image = np.expand_dims(
image, axis=-1
) # Expand dimensions to match model input shape
image = np.expand_dims(image, axis=0) # Expand to add batch dimension
return image
except Exception as e:
print(f"Error processing image: {e}")
return None
# Function to load embedding model and calculate embeddings
def calculate_embeddings(image, model_path="embedding_modelv2.keras"):
# Load the trained model
model = load_model(model_path)
# Remove the final classification layer to get embeddings
embedding_model = Model(inputs=model.input, outputs=model.output)
# Preprocess the image
preprocessed_image = preprocess_image(image)
# Calculate embeddings
embeddings = embedding_model.predict(preprocessed_image)
return embeddings
def classify_image(image):
embeddings = calculate_embeddings(image)
# Convert to 2D array for model input
probabilities = mean_clf.predict_proba(embeddings)[0]
labels = ["Photo", "CGI"]
return {f"{labels[i]}": prob for i, prob in enumerate(probabilities)}
interface = gr.Interface(
fn=classify_image, inputs=["image"], outputs=gr.Label(num_top_classes=2)
)
interface.launch(share=True)