Effective and Efficient Video Compression by the Deep Learning Techniques

Abstract

Deep learning has reached many successes in Video Processing. Video has become a growing important part of our daily digital interactions. The advancement of better resolution content and the large volume offers serious challenges to the goal of receiving, distributing, compressing and revealing highquality video content. In this paper we propose a novel Effective and Efficient video compression by the Deep Learning framework based on the flask, which creatively combines the Deep Learning Techniques on Convolutional Neural Networks (CNN) and Generative Adversarial Networks (GAN). The video compression method involves the layers are divided into different groups for data processing, using CNN to remove the duplicate frames, repeating the single image instead of the duplicate images by recognizing and detecting minute changes using GAN and recorded with Long Short-Term Memory (LSTM). Instead of the complete image, the small changes generated using GAN are substituted, which helps with frame-level compression. Pixel wise comparison is performed using K-nearest Neighbours (KNN) over the frame, clustered with K-means and Singular Value Decomposition (SVD) is applied for every frame in the video for all three colour channels [Red, Green, Blue] to decrease the dimension of the utility matrix [R, G, B] by extracting its latent factors. Video frames are packed with parameters with the aid of a codec and converted to video format and the results are compared with the original video. Repeated experiments on several videos with different sizes, duration, Frames per second (FPS), and quality results demonstrated a significant resampling rate. On normal, the outcome delivered had around a 10% deviation in quality and over half in size when contrasted, and the original video.

Machine Learning Analysis of Enhanced Biodegradable Phoenix dactylifera L./HDPE Composite Thermograms

“…For all samples at different heating rates, Figure 1 and Figure 2 show that the thermograms were non-linear inverse with increasing degradation temperature, indicating the need for the formulation of learning algorithms for the experimental data to be trained using deep neural networks (DNN). A non-linear relationship between input and output data from a system can be more reliable when using deep neural networks, as reported in [ 22 , 23 , 24 ]. Its convolution tends to improve the process as well as reduce the difference between predicted and real data errors.…”

“…Sirca et al [ 22 ] referred to these synaptic weights and biases as arbitrary constants that often change during training. Another article by Panneerselvam [ 24 ] stated that changes in bias during the training process are a sign that the flexibility in the output response creates an offset. Inputs have an impact on outputs, whereas changes to synaptic weights are an indication of the inputs’ influence.…”

“…There are other non-linear activation functions, such as hyperbolic tangents (TanH), rectified linear units (ReLU), exponential linear units (ELU), etc. However, the choice of Sigmoid activation functions is based on their ability to converge data points that are between 0 and 1 based on changes in arbitrary constants during training [ 20 , 24 , 26 ]. For this reason, the input and output data points were divided by the maximum value possible for a TGA experiment as part of data preparation before training.…”

“…More importantly, the ongoing changes to synaptic weights ( ) during the training period contribute to the effects passed on from preceding neurons to succeeding neurons. According to Panneerselvam [ 24 ], a positive value in a trained synaptic weight indicates that the input data directly influences the output signals. Conversely, an increase in negative value in the synaptic weight reduces the likelihood of the input data positively influencing the output signal.…”

“…Reducing this true or residual error to a near-zero level may require an increase in training time or an expansion of the training data capacity, commonly known as data augmentation [ 24 ], to mitigate data overfitting. It is important to note that the DNN framework shown in Figure 5 is characterized by eight hidden neurons (HNS), making it a more complex framework compared to the DNN architecture with six hidden neurons ( Figure 3 ).…”