DeepFake Detector

ANALYZE IMAGE

DROP YOUR IMAGE HERE

or click to browse files

Supported: JPG, PNG, GIF • Max 16MB

Processing with AI...

Confidence Score

DETECTION SAMPLES

Click any sample to test the detector

Waiting for analysis...

Waiting for analysis...

Waiting for analysis...

Waiting for analysis...

PROJECT DEMONSTRATION

Neural Architecture

Built on InceptionV3, our model leverages advanced convolutional neural networks with 23.8M parameters, achieving 97.4% accuracy through transfer learning on 140k+ facial images.

TensorFlow Keras InceptionV3

Optimized Performance

Model trained with GPU acceleration achieving sub-second inference times. Deployed with CPU optimization for web accessibility while maintaining high accuracy and responsive performance for real-time detection.

TensorFlow Flask Web Optimized

Performance Metrics

Exceptional detection capabilities with 97.4% accuracy, 0.067 loss, and robust generalization. Outperforms ResNet50 and matches DenseNet121 with fewer parameters.

Precision: 97% Recall: 98% F1: 97.5%

model_inference.py

                    # InceptionV3 Model Architecture
                    def get_inception_model():
                        inception = InceptionV3(weights='imagenet', include_top=False)
                        model = tf.keras.Sequential([
                            inception,
                            GlobalAveragePooling2D(),
                            Dense(512, activation='relu'),
                            BatchNormalization(),
                            Dropout(0.3),
                            Dense(1, activation='sigmoid')
                        ])
                        return model
                

DROP YOUR IMAGE ANYWHERE

ANALYZE IMAGE

DROP YOUR IMAGE HERE

DETECTION SAMPLES

PROJECT DEMONSTRATION

Neural Architecture

Optimized Performance

Performance Metrics