AnimeCeleb: Large-Scale Animation CelebHeads Dataset for Head Reenactment

Kim, Kangyeol; Park, Sunghyun; Lee, Jae‐Seong; Chung, Sunghyo; Lee, Junsoo; Choo, Jaegul

doi:10.1007/978-3-031-20074-8_24

Cited by 9 publications

(5 citation statements)

References 17 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…PAniC-3D is composed of two major components (Fig. 1a): a 3D reconstructor directly supervised to predict a radiance field from a given front render, and an illustration- [20], AnimePortraits [3], iCartoonFace [43], BizarrePose [7].…”

Section: Methodsmentioning

confidence: 99%

“…It is a fairly common task for 3D character artists to produce a 3D model from a portrait illustration; however from the computer graphics standpoint, this stylized reconstruction setup adds additional ambiguity to an already ill-posed problem. In addition, while there's work in the popular anime/manga domain using 3D character assets (for pose estimation [18], re-targetting [17,20], and reposing [22], etc. ), there's a lack of publicly-available 3D character assets with multi-view renders that allow scalable training (Tab.…”

Section: Anime-style 3d Avatars and Illustrationsmentioning

confidence: 99%

“…We collect a large dataset of 11.2k 3D anime characters from VroidHub, in order to train both the reconstruction and image-to-image translation modules of PAniC-3D. Unlike previous work in the 3D anime domain using MikuMiku-Dance PMD/PMX models [17,18,20], Vroid VRM models conform to the GLTF2 standard [15] with several extensions [37], allowing us to render using ModernGL [10,12]. As the data is crowd-sourced, we filter against a variety of undesirable properties, such as: texture corruption, too much or too little transparency, extreme character sizes, missing bones, etc.…”

Section: Vroid 3d Datasetmentioning

confidence: 99%

See 2 more Smart Citations

Enterprises Build Sustainable Innovation Loop: Take ByteDance Launching Lark as an Example

Chen

2023

Frontiers in Artificial Intelligence and Applications

View full text Add to dashboard Cite

This paper uses the method of formal model construction to model the path of enterprise sustainable innovation and explains it by case analysis. Some suggestions are put forward then. In the changing market environment, enterprises can adapt to and strive for more living space only through continuous innovation. The construction of a sustainable innovation loop makes the previous research on sustainable innovation become relatively concrete from the abstract. Enterprises should construct and complete the loop of innovation path according to their own characteristics. In the process of launching Lark, ByteDance takes scientific and technological innovation as the basis, and business model innovation as the key link, and finally completes the cycle of innovation path, and the innovation is still ongoing, achieving the purpose of sustainable innovation, reflecting the enterprise innovation strategy and corporate culture.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Anime-style 3d Avatars and Illustrationsmentioning

confidence: 99%

Section: Vroid 3d Datasetmentioning

confidence: 99%

See 1 more Smart Citation

Enterprises Build Sustainable Innovation Loop: Take ByteDance Launching Lark as an Example

Chen

2023

Frontiers in Artificial Intelligence and Applications

View full text Add to dashboard Cite

show abstract

“…Human-centric synthetic datasets: Various synthetic human body datasets [20,22,49,67] are proposed in the literature. However, they are far less developed compared to artificial human face datasets [23,87] these datasets face the problems of unnatural interactions between background and humans and the lack of character diversities. For example, the anime/manga character dataset in [22] contains only 2631 different characters despite having more than a million images.…”

Section: Related Workmentioning

confidence: 99%

Human-Art: A Versatile Human-Centric Dataset Bridging Natural and Artificial Scenes

Ju¹,

Zeng²,

Wang³

et al. 2023

Preprint

View full text Add to dashboard Cite

Humans have long been recorded in a variety of forms since antiquity. For example, sculptures and paintings were the primary media for depicting human beings before the invention of cameras. However, most current human-centric computer vision tasks like human pose estimation and human image generation focus exclusively on natural images in the real world. Artificial humans, such as those in sculptures, paintings, and cartoons, are commonly neglected, making existing models fail in these scenarios.As an abstraction of life, art incorporates humans in both natural and artificial scenes. We take advantage of it and introduce the Human-Art dataset to bridge related tasks in natural and artificial scenarios. Specifically, Human-Art contains 50k high-quality images with over 123k person instances from 5 natural and 15 artificial scenarios, which are annotated with bounding boxes, keypoints, self-contact points, and text information for humans represented in both 2D and 3D. It is, therefore, comprehensive and versatile for various downstream tasks. We also provide a rich set of baseline results and detailed analyses for related tasks, including human detection, 2D and 3D human pose estimation, image generation, and motion transfer. As a challenging dataset, we hope Human-Art can provide insights for relevant research and open up new research questions.

show abstract

“…Trinity [67] 2018 224 (Min) Videos-Text-Audio-Skelton(2,3D) English Link TED-Gesture [68] 2019 ∼252k Videos-Text-Audio-Skelton(2D) English Link Talking With Hands [69] 2019 200 Videos-Text-Audio-Skelton(2,3D) English Link Speech2Gesture [70] 2019 ∼60k Videos-Text-Audio-Skelton(2D) English Link TED-Expressive [27] 2022 ∼252k Videos-Text-Audio-Skelton(2,3D) English Link Talking Head GRID [93] 2006 ∼34k Video-Text English Link eNTERFACE [94] 2006 ∼1k Video-Text-Audio Multiple Link MIRACL-VC1 [95] 2014 ∼3k Video-Text-Depth English Link CREMA-D [96] 2015 ∼7k Video-Text-Audio English Link TCD-TIMIT [96] 2015 ∼7k Video-Text-Audio English Link MODALITY [97] 2015 ∼6k Video-Text-Audio English Link LRW [98] 2016 ∼539k Video-Text English Link MSP-IMPROV [99] 2016 ∼ 1K Video-Text-Audio English Link ObamaSet [100] 2017 ∼1k Video-Text-Audio English Link VoxCeleb1 [101] 2017 ∼22k Video-Text-Audio English Link VoxCeleb2 [102] 2018 ∼146k Video-Text-Audio English Link LRS2 [103] 2018 ∼96k Video-Text English Link LRS3-TED [104] 2018 ∼119k Video-Text English Link RAVDESS [105] 2018 ∼1k Video-Text-Audio English Link MELD [106] 2018 ∼13k Video-Text-Audio English Link AVSpeech [107] 2018 ∼150k Video-Audio Multiple Link VOCASET [108] 2019 480 Video-Text-Audio-3DFace English Link LRW-1000 [109] 2019 ∼718K Video-Text Chinese Link FaceForensics++ [110] 2019 ∼1k Video-Text-Audio English Link MEAD [111] 2020 ∼281k Video-Text-Audio English Link HDTF [112] 2021 ∼10k Video-Text-Audio English Link AnimeCeleb [113] 2022 ∼2.4M Video-Text-Audio-3DFace English Link VLRDT [114] 2022 ∼2k Video-Text Turkish Link KoEBA [115] 2023 104(P) Video-Text-Audio Korea Link…”

Section: Co-speechmentioning

confidence: 99%

Multi-Modal Fusion Emotion Recognition Method of Speech Expression Based on Deep Learning

Liu

Wang

et al. 2021

Front. Neurorobot.

View full text Add to dashboard Cite

The redundant information, noise data generated in the process of single-modal feature extraction, and traditional learning algorithms are difficult to obtain ideal recognition performance. A multi-modal fusion emotion recognition method for speech expressions based on deep learning is proposed. Firstly, the corresponding feature extraction methods are set up for different single modalities. Among them, the voice uses the convolutional neural network-long and short term memory (CNN-LSTM) network, and the facial expression in the video uses the Inception-Res Net-v2 network to extract the feature data. Then, long and short term memory (LSTM) is used to capture the correlation between different modalities and within the modalities. After the feature selection process of the chi-square test, the single modalities are spliced to obtain a unified fusion feature. Finally, the fusion data features output by LSTM are used as the input of the classifier LIBSVM to realize the final emotion recognition. The experimental results show that the recognition accuracy of the proposed method on the MOSI and MELD datasets are 87.56 and 90.06%, respectively, which are better than other comparison methods. It has laid a certain theoretical foundation for the application of multimodal fusion in emotion recognition.

show abstract

AnimeCeleb: Large-Scale Animation CelebHeads Dataset for Head Reenactment

Cited by 9 publications

References 17 publications

Enterprises Build Sustainable Innovation Loop: Take ByteDance Launching Lark as an Example

Enterprises Build Sustainable Innovation Loop: Take ByteDance Launching Lark as an Example

Human-Art: A Versatile Human-Centric Dataset Bridging Natural and Artificial Scenes

Multi-Modal Fusion Emotion Recognition Method of Speech Expression Based on Deep Learning

Contact Info

Product

Resources

About