Scenic

Scenic is a project that utilizes transfer learning with a pretrained ResNet model using TensorFlow/Keras for classifying natural scene images. The project leverages the Intel Image Classification Dataset for training and evaluation.

Features

• Built using TensorFlow/Keras.
• Employs transfer learning with pretrained ResNet weights for efficient feature extraction.
• Classifies natural scenes into six categories: buildings, forest, glacier, mountain, sea, and street.

Dataset

• Intel Image Classification Dataset
  The dataset contains natural scene images divided into six categories.

  Dataset source: Intel Image Classification Dataset

Results

• ResNet50 with transfer learning achieved 92% accuracy on the test dataset.

License

This project is licensed under the MIT License. See the LICENSE file for more details.

Acknowledgments

• Dataset by Intel.
• Pretrained ResNet weights from Keras Applications.
• ResNet architecture: "Deep Residual Learning for Image Recognition".