Please enable JavaScript.
Coggle requires JavaScript to display documents.
The Path of Film and Television Animation Creation Using Virtual Reality…
The Path of Film and Television Animation Creation Using Virtual Reality Technology under the Artificial
Intelligence
Virtual Reality
Abilities
perception
interactive feedback
gesture interaction
static gesture recognition
changed to
convolutional neural network (CNN)
has more efficient recognition ability
gesture recognition is classifier
to use edge detection to obtain the gesture features
CNN-SVM support vector machine
1 more item...
contour descriptor based on the depth projection map
to obtain the hand shape
to obtain structure information in depth image
support vector machine (SVM) classification
The recognition accuracy has been improved
dynamic gesture recognition
Main function
to create a simulated simulation environment to achieve very realistic effects like real life
an image in a three-dimensional (3D) space
interactive feedback
vision
hearing
touch
orientation
records and analyzes
people’s relevant
actions
physical activity data
sound in a three-dimensional (3D) space
the ability of online perception
Making animation
enriched
at the perceptual level
added interactive mode
watching animation to participate in the development of the plot
traditional form of animation won't disappear from people’s vision because of the VR technology
many traditional animations launched in Japan have achieved strong influence in multiple countries
The development path of VR animation
The transformation from “plane” to “stereo” vision:
The production of traditional animation is hand-drawn by relevant workers.
After the computer appeared, the computer monitor replaced the paper.
However, it is still difficult to show a 3D feeling (computer screen)
VR gets rid of the previous screen and creates a 3D and realistic space
Narrative transformation from “linear” to “branch”:
VR technology can reflect the feedback ability of animation
the “linear” development structure
the beginning
development
climax
end of things
The audience is more passive to accept the whole story and has no impact on the development of the whole story
VR technology itself has the characteristics of interactivity, which has brought different narrative forms to the development of the whole story.
The transformation from “watching” to “being present”:
the audience, as an independent individual, is a “bystander,
has no relationship with any element in the animation
no matter how optimized the environment is, it cannot be denied that the audience is still a bystander
VR technology enable the audience to participate in the interaction in the animation from the first perspective
from “watching” to “being present.”
SVM Classifier
uses different kernel functions
transform the samples that cannot be divided into low-dimensional input space into high-dimensional feature space
theoretical basis
to minimize institutional risk, form the best hyperplane in the feature space
to obtain the structured information of data distribution
to reduce the error of independent test set
LIBSVM
Developed by Chih-jen Lin
used to build SVM
for efficient classification and regression
can solve multiple classification problems
When making classification decisions, LIBSVM uses the maximum winning algorithm
Each classifier will vote on the category it determines, and the final classification result will be qualified by the category with a higher number of votes.
CNN Classifier
deep feedforward neural network
has two parts
an automatic feature extractor
trainable classifier
can make the deep CNN structure automatically obtain the high-level features of the image
Caffe framework
adopted to build CNN
Krizhevsky network (AlexNet)
as the training network model
CNN-SVM Hybrid Model
is to replace the last output layer in CNN with SVM
Process:
The raw image is sent to the input layer of the CNN for training until convergence or enough iterations.
The training samples' images are sent to the trained CNN model, yielding 2048-dimensional samples.
training set
The obtained sample feature vecto
to train the SVM classifier
to obtain the CNN-SVM hybrid model
LIBSVM
estimates the probability that each sample is divided into a certain category
problem of N classification
distance
The similarity between the two classes
The smaller the distance is, the smaller the gap between the two is
The absolute value of the prediction probability difference
Experiment
Date
Relevant gesture images of 400 college students
left hands 2 meters
in front of Kinect
Total
4000 depth images
will not be disturbed by the environment
Light
will preserve the structural features of the human hand
segmented
to reduce the background interference of the color image
Process
stored as a grayscale depth image with a pixel value of [0–255]
the gray value 125 is used as the threshold
defined as a mask image
skin color segmentation
1 more item...
4000 color images
It is still difficult to recognize gestures accurately
why?
complex background conditions
appearance
shape
Case Analysis
the gesture image will be segmented into 30000 images
26000 images for model training
4000 images for testing
Tested models
AlexNet
optimized CNN-SVM
CNN-SVM
Results
the optimized CNN-SVM is higher than that of the original CNN-SVM
the optimized CNN-SVM has a recognition accuracy of 0.97
Conclusion
the optimized CNN-SVM plays an important role in improving the ability of interactive feedback of VR
the optimized CNN-SVM enhances the interactive ability of films and television animation works
Source:
https://www.hindawi.com/journals/sp/2022/1712929/
Definitions
distribution, convergence, hyperplane, grayscale depth, SVM, LIBSVM, CNN, classifier, institutional risk, interactive feedback
Support Vector Machine - a supervised machine learning algorithm used for classification and regression tasks.
A library for SVMs. It provides various SVM implementations for classification, regression, and distribution estimation tasks.
Convolutional Neural Network - a deep learning algorithm commonly used for analyzing visual imagery.
A machine learning model that assigns a label or category to input data based on its features.
Refers to the potential losses or adverse impacts faced by an institution, such as a bank, corporation, or government entity
A process where a system or user interface provides immediate responses or suggestions based on the user's actions or input.
Refers to the set of all possible values and their corresponding probabilities or frequencies of occurrence.
A subspace of one dimension less than its ambient space.
Refers to the property of a sequence of numbers, functions, or algorithms to approach a certain value, usually a solution or an equilibrium point.
Refers to the number of bits used to represent the color of each pixel in a grayscale image.
