An area not yet fully explored - cartoon face recognition

As one of the most representative technologies for deep learning, face recognition has been rapidly developed in recent years. It has greatly improved in accuracy, speed, and scale. While achieving such great success, scientists have turned their attention to an area that has not yet been fully explored - cartoon face recognition, which will accurately detect and identify faces in cartoon animation works.

With the rapid development of digital technology and new media technologies, more and more cartoon characters have appeared in our lives. They not only become children's windows for understanding the world, but also provide teaching aids and scientific research dissemination functions, even with To describe personal opinions and even spread social thoughts. Compared with standard paintings, cartoons, comics, and humorous works are expressed in a very exaggerated manner, and have led to large-scale deformation and transfer of features. The demand for cartoon face detection in the animation industry is also increasing day by day, mainly including: searching for similar cartoon images in the network through an image search engine; helping visual impairments recognize cartoon movie fun through recognition and speech synthesis; and Social media content is processed as part of content control and review software.

In order to achieve this goal, the researchers used the IIIT-CFW cartoon data set as a fuel, based on deep learning to achieve a cartoon face detection, identification work, and exceeded the traditional method to achieve excellent detection results.

This data set contains 8928 tagged cartoon face images, which not only include basic face data of different ages, genders, and emotions, but also include high dimensional information such as various races, face positions, and sarcastic metaphors. Examples of tagging information include the following seven feature dimensions and five face position dimension information including the character's name:

Face Annotation

The author conducted three main tasks for the cartoon face: face detection, face recognition and gender detection, and identified and extracted the key points of the cartoon face during the detection process.

For face detection, the researchers mainly use the MTCNN (Multi-task Cascaded Convolutional Network) network. This architecture consists of three main parts: the proposed network P-Net, which proposes candidate frames through the image pyramid, and is subsequently used. The refinement network R-Net for refining the optimization results is finally the output network O-Net for generating the final face frame and the results of the five flags.

For face recognition work, the researchers proposed two methods. The first is to use Inceptionv3+SVM for recognition. First, use Inception to efficiently extract image features, and finally use a classifier to classify the final 2048-dimensional results. Identification.

Another method uses a suggestion system to construct a marker extraction system and realizes face recognition based on a multi-input multi-output CNN classifier. First, the cartoon image is grayscaled and normalized, and then the coordinates of the fifteen feature points of the cartoon face image are extracted. In the detection process, a 5-layer neural network was used to detect feature points (training was performed using real human faces to improve face feature detection capabilities). The final result and the pixel processing result are sent to the proposed face recognition architecture network at the same time, and the CNN multi-input and multi-output results are used for identification.

After training on the above network, good test results were obtained. In the performance of face detection, the author also compared the results based on HOG features and Harr features. The indicators of true positives, false positives, and false negatives surpass these methods.

It also performs well in tasks extracted from cartoon faces.

The main contribution of this work is to use the MTCNN framework to achieve the detection of cartoon faces, and has greatly improved on multiple indicators. At the same time, the pre-trained Inception framework and SVM feature classification are used in cartoon face recognition tasks. The implement achieves higher effect. And proposed LeNet-based multi-input multi-output HCNN architecture reduces the top5 error rate.

It is worth mentioning that in the cartoon image field there is still a lot of work to fill in pits, including cartoon face recognition, verification, gender recognition, photo to cartoon image conversion (similar to style migration), cartoon face detection, location Estimation and keypoint detection, correlation feature recognition, and search engines based on cartoon image features. It also includes the generation, sketching and rendering of a series of cartoon expressions through photos, and the corresponding inverse problems. There are many applications and points that can be explored in the cartoon field. Even GANs and VAE are a good choice!

Mini Gaming PC


What suddenly appear into your mind when hear Mini Gaming PC? Is there quality heat-releasing fan and box design of Mini Gaming PC Build? Does it run stable and smoothly when handle heavier jobs, like Photoshop, Pr, engineering design and drawing, 3d Max, big games, etc. ? now we can be responsible to answer [ Yes". No matter cheap mini gaming pc or Mini Gaming PC Under $500, equips with quality fan to heat releasing. Therefore, no need to worry that again.

To processor, can do from intel celeron j4125, N5105 up to i3, i5 i7 10th 11th 12th with or without video graphics. To slots, same rich as traditional computer tower, like above 4 or 6 usb ports, 2 Rj45, VGA, PD, slot, etc.

You can also find Mini Gaming PC, j4125 Mini PC ,Custom All In One PC, windows 10 Education Laptop,Quad Core Processor Laptop, 15 Inch Gaming Laptop ,windows Yoga Laptop , 8 inch Android Tablet, etc.

If any other special requirements interest, you can also let us know, will try our best support you.

Meet your unique demand in this field is our mission, so just feel free contact us whenever you have different idea.


Mini Gaming PC Build,Cheap Mini Gaming PC,Mini Gaming PC Under $500,Mini Gaming Desktop

Henan Shuyi Electronics Co., Ltd. , https://www.shuyioemelectronics.com

Posted on