VideoNetClassification

Table of Contents

• Introduction
• Collaborators
• Installation
• Usage
• Project Structure
• Data Loading
• Model Training
• Evaluation
• License

Introduction

VideoNetClassification is a project focused on classifying videos using a Recurrent Neural Network (RNN). This project leverages TensorFlow and TensorFlow Hub for model building and training. It uses the UCF101 dataset, a popular action recognition dataset.

Collaborators

• Yahia Ehab
• Mariam Amr
• Mohamed Khaled

Installation

To install the required dependencies, you can use the following commands:

pip install imageio
pip install opencv-python
pip install git+https://github.com/tensorflow/docs

Usage

1. Clone the repository.
2. Install the required dependencies as mentioned above.
3. Run the notebook.ipynb notebook to train and evaluate the model.

Project Structure

• notebook.ipynb: The main Jupyter Notebook containing all the code for data loading, preprocessing, model training, and evaluation.

Data Loading

The UCF101 dataset is used in this project. The dataset is loaded and preprocessed using various helper functions. Videos are downloaded and their frames are extracted and resized.

Model Training

An RNN model is created using Keras. The training process involves:

1. Extracting features from video frames using a pre-trained CNN.
2. Training the RNN model with the extracted features.

Steps:

1. Feature Extraction:

   base_model = tf.keras.applications.InceptionV3(weights='imagenet', include_top=False, pooling='avg')
   video_features = []
   for video in video_frames:
       video_features.append(base_model.predict(video))

2. Data Preparation:

   split = int(0.8 * len(video_features))
   train, test = video_features[:split], video_features[split:]
   train_labels, test_labels = labels[:split], labels[split:]

3. Label Encoding:

   from tensorflow.keras.utils import to_categorical
   from sklearn.preprocessing import LabelEncoder
   
   label_encoder = LabelEncoder()
   train_labels_encoded = label_encoder.fit_transform(train_labels)
   train_labels_onehot = to_categorical(train_labels_encoded, num_of_classes)
   
   test_labels_encoded = label_encoder.fit_transform(test_labels)
   test_labels_onehot = to_categorical(test_labels_encoded, num_of_classes)

4. Model Creation and Training:

   from keras.models import Sequential
   from keras.layers import SimpleRNN, Dense
   
   rnn_model = Sequential()
   rnn_model.add(SimpleRNN(50, input_shape=(sequence_length, 2048)))
   rnn_model.add(Dense(num_of_classes, activation="softmax"))
   
   rnn_model.compile(loss="categorical_crossentropy", optimizer="adam", metrics=["accuracy"])
   
   rnn_model.fit(x=train, y=train_labels_onehot, epochs=15, batch_size=64, validation_split=0.2)

Evaluation

The trained model is evaluated using the test dataset:

evaluation = rnn_model.evaluate(x=test, y=test_labels_onehot)
print("Evaluation results:", evaluation)

License

This project was given and managed by the GIU