正在加载视频...

视频加载失败

加载此视频时出现问题。这可能是由于临时网络问题，或视频可能不可用。

4D LangSplat: 4D Language Gaussian Splatting via Multimodal Large Language Models Contributions: • We introduce 4D LangSplat for open-vocabulary 4D spatial-temporal queries. To the best of our knowledge, we are the first to construct 4D language fields with object textual captions generated by MLLMs. • To model smooth transitions... show more

MrNeRF

16,826 subscribers

10,961 次观看 • 1 年前 •via X (Twitter)

教育科学技术

Anya Rossi• Live Now

Private livecam show

5 条评论

MrNeRF 的头像

MrNeRF1 年前

Paper: Project: YouTube: Code:

AssemblyAI 的头像

AssemblyAI1 年前

Announcing: Our most advanced speech-to-text model goes beyond accuracy to capture the real-world complexity of human conversation and deliver reliable, source-of-truth audio data. Explore Universal-2 updates 👇

GifCo 的头像

GifCo1 年前

Can think of a lot of cool use cases for this!

MrNeRF 的头像

MrNeRF1 年前

Can you share some ideas? I am curious :)

LLMLens 的头像

LLMLens1 年前

4D LangSplat's fusion of spatiotemporal Gaussian splatting with LLMs echoes Simondon's concept of technical individuation. Yet it risks reifying language as mere technical object, divorced from lived experience. How might we preserve the human in this hyper-technical assemblage?

相关视频

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Robot Learning needs 4D world models! Robot Learning needs 4D world models! Robot Learning needs 4D world models! We introduce TesserAct, a 4D embodied world model that can simulate how agents interact with the 3D world over time! We achieve this by simply extending a pre-trained 2D video generation model to jointly predict RGB, depth, and surface normals. It enables: 1️⃣ Much better policy learning in the wild 2️⃣ Temporal + spatial coherence in 4D dynamic prediction 3️⃣ Novel view synthesis for embodied scenes Code: Paper Link: Project page:

Chuang Gan

43,265 次观看 • 1 年前

1000+ FPS 4D Gaussian Splatting for Dynamic Scene Rendering Contributions: • We delve into the temporal redundancy of 4D Gaussian Splatting and explain the main reason for the storage pressure and suboptimal rendering speed. • We introduce 4DGS-1K, a compact and memory-efficient framework to address these issues. It consists of two key components: a spatial-temporal variation score-based pruning strategy and a temporal filter. • Extensive experiments demonstrate that 4DGS-1K not only achieves a substantial storage reduction of approximately 41× but also accelerates rasterization to 1000+ FPS while maintaining high-quality reconstruction.

1000+ FPS 4D Gaussian Splatting for Dynamic Scene Rendering Contributions: • We delve into the temporal redundancy of 4D Gaussian Splatting and explain the main reason for the storage pressure and suboptimal rendering speed. • We introduce 4DGS-1K, a compact and memory-efficient framework to address these issues. It consists of two key components: a spatial-temporal variation score-based pruning strategy and a temporal filter. • Extensive experiments demonstrate that 4DGS-1K not only achieves a substantial storage reduction of approximately 41× but also accelerates rasterization to 1000+ FPS while maintaining high-quality reconstruction.

MrNeRF

12,366 次观看 • 1 年前

We’ve upgraded Stable Video Diffusion 4D to Stable Video 4D 2.0 (SV4D 2.0), improving the quality of 4D outputs generated from a single object-centric video. While 3D provides a static view of an object’s shape and size; 4D extends this by including time, showing how the object moves. This multi-view video diffusion model generates a 4D output in three steps: 1️⃣ Starts with an input video of a moving person or object 2️⃣ Generates novel views of the subject from unseen angles 3️⃣ Constructs a single dynamic 4D video output with spatial and temporal consistency You can learn more here: (1/4)

We’ve upgraded Stable Video Diffusion 4D to Stable Video 4D 2.0 (SV4D 2.0), improving the quality of 4D outputs generated from a single object-centric video. While 3D provides a static view of an object’s shape and size; 4D extends this by including time, showing how the object moves. This multi-view video diffusion model generates a 4D output in three steps: 1️⃣ Starts with an input video of a moving person or object 2️⃣ Generates novel views of the subject from unseen angles 3️⃣ Constructs a single dynamic 4D video output with spatial and temporal consistency You can learn more here: (1/4)

Stability AI

36,041 次观看 • 1 年前

Meet our inhouse tech Doggo "Rolo". Cinematic RGB Lighting "More Doggo than Doggo" Since July we've been redesigning our scanning pipeline to work with the amazing 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

Meet our inhouse tech Doggo "Rolo". Cinematic RGB Lighting "More Doggo than Doggo" Since July we've been redesigning our scanning pipeline to work with the amazing 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

Infinite-Realities

98,702 次观看 • 3 年前

White Light Reference for Machine Learning. Meet our inhouse tech Doggo "Rolo". "More Doggo than Doggo" Since July we've been redesigning our scanning pipeline to work with the new 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come.. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

White Light Reference for Machine Learning. Meet our inhouse tech Doggo "Rolo". "More Doggo than Doggo" Since July we've been redesigning our scanning pipeline to work with the new 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come.. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

Infinite-Realities

236,837 次观看 • 3 年前

"More Doggo than Doggo" Motion scanned @ 60fps. Playing back @ 20fps. Meet our inhouse tech Doggo "Rolo". Since July we've been redesigning our scanning pipeline to work with the new 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. Our next goal is to optimize the player to be multi-threaded to improve performance over 60fps. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

"More Doggo than Doggo" Motion scanned @ 60fps. Playing back @ 20fps. Meet our inhouse tech Doggo "Rolo". Since July we've been redesigning our scanning pipeline to work with the new 3D Gaussian Splatting for Real-Time Radiance Field Rendering method from Inria. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria IR's AeonX capture system has been designed to capture "Spatial" memories for the future. We've built on a custom version of Inria's Sibr viewer to handle the playback of our multiple processed 4D frames. Our next goal is to optimize the player to be multi-threaded to improve performance over 60fps. GS has reignited our passion for 3D/4D scanning. Thank you to the team at Inria More content to come. #guassiansplatting #ir #aeonx #4d #motionscanning #inria #ximea #idatronic #spatial #youth83

Infinite-Realities

48,330 次观看 • 3 年前

LangSplatV2: High-dimensional 3D Language Gaussian Splatting with 450+ FPS Contributions: • LangSplatV2 achieves real-time performance with 476.2 FPS for high-dimensional feature splatting and 384.6 FPS for 3D open-vocabulary text querying. • Delivers a 42× speedup and 47× boost in performance compared to LangSplat. • Improves query accuracy while drastically reducing inference time. • Replaces the heavyweight decoder in LangSplat with a sparse coefficient field, removing the main performance bottleneck. • Introduces a CUDA-optimized sparse coefficient splatting method, enabling fast and high-quality rendering of high-dimensional features. • Enables scalable 3D language interaction in complex scenes, opening up real-time applications previously not possible with LangSplat.

LangSplatV2: High-dimensional 3D Language Gaussian Splatting with 450+ FPS Contributions: • LangSplatV2 achieves real-time performance with 476.2 FPS for high-dimensional feature splatting and 384.6 FPS for 3D open-vocabulary text querying. • Delivers a 42× speedup and 47× boost in performance compared to LangSplat. • Improves query accuracy while drastically reducing inference time. • Replaces the heavyweight decoder in LangSplat with a sparse coefficient field, removing the main performance bottleneck. • Introduces a CUDA-optimized sparse coefficient splatting method, enabling fast and high-quality rendering of high-dimensional features. • Enables scalable 3D language interaction in complex scenes, opening up real-time applications previously not possible with LangSplat.

MrNeRF

14,261 次观看 • 1 年前

4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos Abstract: We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos. Using 4D Gaussian as an inductive bias, 4DGT unifies static and dynamic components, enabling the modeling of complex, time-varying environments with varying object lifespans. We introduced a novel density control strategy in training, which allows our 4DGT to handle longer space-time input while maintaining efficient rendering at runtime. Our model processes 64 consecutive posed frames in a rolling-window fashion, predicting consistent 4D Gaussians in the scene. Unlike optimization-based methods, 4DGT performs purely feed-forward inference, reducing reconstruction time from hours to seconds and scaling effectively to long video sequences. Trained only on large-scale monocular posed video datasets, 4DGT can significantly outperform prior Gaussian-based networks in real-world videos and achieve on-par accuracy with optimization-based methods on cross-domain videos.

4DGT: Learning a 4D Gaussian Transformer Using Real-World Monocular Videos Abstract: We propose 4DGT, a 4D Gaussian-based Transformer model for dynamic scene reconstruction, trained entirely on real-world monocular posed videos. Using 4D Gaussian as an inductive bias, 4DGT unifies static and dynamic components, enabling the modeling of complex, time-varying environments with varying object lifespans. We introduced a novel density control strategy in training, which allows our 4DGT to handle longer space-time input while maintaining efficient rendering at runtime. Our model processes 64 consecutive posed frames in a rolling-window fashion, predicting consistent 4D Gaussians in the scene. Unlike optimization-based methods, 4DGT performs purely feed-forward inference, reducing reconstruction time from hours to seconds and scaling effectively to long video sequences. Trained only on large-scale monocular posed video datasets, 4DGT can significantly outperform prior Gaussian-based networks in real-world videos and achieve on-par accuracy with optimization-based methods on cross-domain videos.

MrNeRF

34,782 次观看 • 1 年前

[NeurIPS '24] DreamMesh4D: Video-to-4D Generation with Sparse-Controlled Gaussian-Mesh Hybrid Representation Abstract (excerpt) We introduce DreamMesh4D, a novel framework that combines mesh representation with sparse-controlled deformation technique to generate high-quality 4D object from a monocular video. To overcome the limitation of classical texture representation, we bind Gaussian splats to the surface of the triangular mesh for differentiable optimization of both the texture and mesh vertices. In particular, DreamMesh4D begins with a coarse mesh provided by a single image based 3D generation method. Sparse points are then uniformly sampled across the surface of the mesh, and are used to build a deformation graph to drive the motion of the 3D object for the sake of computational efficiency and providing additional constraint. For each step, transformations of sparse control points are predicted using a deformation network, and the mesh vertices as well as the bound surface Gaussians are deformed via a geometric skinning algorithm. The skinning algorithm is a hybrid approach combining LBS (linear blending skinning) and DQS (dual-quaternion skinning), mitigating drawbacks associated with both approaches. The static surface Gaussians and mesh vertices as well as the dynamic deformation network are learned via reference view photometric loss, score distillation loss as well as other regularization losses in a two-stage manner. Extensive experiments demonstrate that our method outperforms prior video-to-4D generation methods in terms of rendering quality and spatial-temporal consistency.

[NeurIPS '24] DreamMesh4D: Video-to-4D Generation with Sparse-Controlled Gaussian-Mesh Hybrid Representation Abstract (excerpt) We introduce DreamMesh4D, a novel framework that combines mesh representation with sparse-controlled deformation technique to generate high-quality 4D object from a monocular video. To overcome the limitation of classical texture representation, we bind Gaussian splats to the surface of the triangular mesh for differentiable optimization of both the texture and mesh vertices. In particular, DreamMesh4D begins with a coarse mesh provided by a single image based 3D generation method. Sparse points are then uniformly sampled across the surface of the mesh, and are used to build a deformation graph to drive the motion of the 3D object for the sake of computational efficiency and providing additional constraint. For each step, transformations of sparse control points are predicted using a deformation network, and the mesh vertices as well as the bound surface Gaussians are deformed via a geometric skinning algorithm. The skinning algorithm is a hybrid approach combining LBS (linear blending skinning) and DQS (dual-quaternion skinning), mitigating drawbacks associated with both approaches. The static surface Gaussians and mesh vertices as well as the dynamic deformation network are learned via reference view photometric loss, score distillation loss as well as other regularization losses in a two-stage manner. Extensive experiments demonstrate that our method outperforms prior video-to-4D generation methods in terms of rendering quality and spatial-temporal consistency.

MrNeRF

12,323 次观看 • 1 年前

Current world models often separate spatial reasoning from temporal modeling, leading to consistency bottlenecks and scaling limits. We developed RAYNOVA, a 4D world foundation model, to address this by unifying space and time into a single representation via a pure auto-regressive framework. The result: Data-driven high-fidelity simulation data that scales across any vehicles, anywhere. Read the full technical breakdown:

Current world models often separate spatial reasoning from temporal modeling, leading to consistency bottlenecks and scaling limits. We developed RAYNOVA, a 4D world foundation model, to address this by unifying space and time into a single representation via a pure auto-regressive framework. The result: Data-driven high-fidelity simulation data that scales across any vehicles, anywhere. Read the full technical breakdown:

Applied Intuition

11,252 次观看 • 3 个月前

4D videos, are we cooked? 😳 uses 4D Gaussian Splatting to reconstruct real footage into fully explorable scenes, letting you move the camera freely through a moment after it’s been filmed. Combine it with tools like Seedance for character replacement and motion control, and Beeble for relighting, and you unlock entire cinematic workflows. You can even step inside your scenes in augmented reality with the Apple Vision Pro. #4DVAI #GaussianSplatting #AIVideoWorkflow #CreativeAI #SpatialVideo

4D videos, are we cooked? 😳 uses 4D Gaussian Splatting to reconstruct real footage into fully explorable scenes, letting you move the camera freely through a moment after it’s been filmed. Combine it with tools like Seedance for character replacement and motion control, and Beeble for relighting, and you unlock entire cinematic workflows. You can even step inside your scenes in augmented reality with the Apple Vision Pro. #4DVAI #GaussianSplatting #AIVideoWorkflow #CreativeAI #SpatialVideo

Rourke Heath

69,296 次观看 • 1 个月前

Check out 𝐌𝐨𝐭𝐢𝐨𝐧𝟐𝐕𝐞𝐜𝐒𝐞𝐭𝐬, a 4D diffusion model for dynamic surface reconstruction from imperfect observations of sparse, noisy, or partial point clouds. Main idea: we represent time-varying shapes via 4D neural representation with latent vector sets, and then explicitly learns the shape and motion distribution of non-rigid objects through an iterative denoising process of latent vector sets. Great work by Wei Cao るちゃん Biao Zhang Jiapeng Tang

Check out 𝐌𝐨𝐭𝐢𝐨𝐧𝟐𝐕𝐞𝐜𝐒𝐞𝐭𝐬, a 4D diffusion model for dynamic surface reconstruction from imperfect observations of sparse, noisy, or partial point clouds. Main idea: we represent time-varying shapes via 4D neural representation with latent vector sets, and then explicitly learns the shape and motion distribution of non-rigid objects through an iterative denoising process of latent vector sets. Great work by Wei Cao るちゃん Biao Zhang Jiapeng Tang

Matthias Niessner

29,142 次观看 • 2 年前

[SIGGRAPH Asia '24 (TOG)] Representing Long Volumetric Video with Temporal Gaussian Hierarchy Contributions: • We introduce a novel, efficient, and expressive Temporal Gaussian Hierarchy representation for long volumetric video. To our knowledge, our method is the first approach capable of handling minutes of volumetric video data. • We propose a Compact Appearance Model and a new rasterization implementation to facilitate real-time, high-quality dynamic view synthesis while maintaining a compact size. • We propose a system to efficiently model long volumetric videos for the first time and demonstrate state-of-the-art dynamic view synthesis quality on the Neural3DV [Li et al. 2022], ENeRF-Outdoor [Lin et al. 2022], and MobileStage [Xu et al. 2024b] datasets, while also achieving the best rendering speed with reduced training cost and memory usage.

[SIGGRAPH Asia '24 (TOG)] Representing Long Volumetric Video with Temporal Gaussian Hierarchy Contributions: • We introduce a novel, efficient, and expressive Temporal Gaussian Hierarchy representation for long volumetric video. To our knowledge, our method is the first approach capable of handling minutes of volumetric video data. • We propose a Compact Appearance Model and a new rasterization implementation to facilitate real-time, high-quality dynamic view synthesis while maintaining a compact size. • We propose a system to efficiently model long volumetric videos for the first time and demonstrate state-of-the-art dynamic view synthesis quality on the Neural3DV [Li et al. 2022], ENeRF-Outdoor [Lin et al. 2022], and MobileStage [Xu et al. 2024b] datasets, while also achieving the best rendering speed with reduced training cost and memory usage.

MrNeRF

79,465 次观看 • 1 年前

SuperGSeg: Open-Vocabulary 3D Segmentation with Structured Super-Gaussians Contributions: • We propose SuperGSeg: a 3D segmentation method with neural Gaussians, designed to learn hierarchical instance segmentation features from 2D foundation models. • We introduce the concept of Super-Gaussian, a novel representation that integrates hierarchical instance segmentation features, enabling the embedding of high-dimensional language features. This approach addresses previously unfeasible challenges in representing complex scenes with rich semantic details. • Extensive experiments on the LERF-OVS and ScanNet datasets demonstrate the effectiveness of the proposed method, achieving significant improvements in open-vocabulary 3D object-level and scene-level semantic segmentation. It shows particular strength in capturing fine-grained scene details and dense pixel semantic segmentation tasks for the first time.

SuperGSeg: Open-Vocabulary 3D Segmentation with Structured Super-Gaussians Contributions: • We propose SuperGSeg: a 3D segmentation method with neural Gaussians, designed to learn hierarchical instance segmentation features from 2D foundation models. • We introduce the concept of Super-Gaussian, a novel representation that integrates hierarchical instance segmentation features, enabling the embedding of high-dimensional language features. This approach addresses previously unfeasible challenges in representing complex scenes with rich semantic details. • Extensive experiments on the LERF-OVS and ScanNet datasets demonstrate the effectiveness of the proposed method, achieving significant improvements in open-vocabulary 3D object-level and scene-level semantic segmentation. It shows particular strength in capturing fine-grained scene details and dense pixel semantic segmentation tasks for the first time.

MrNeRF

13,594 次观看 • 1 年前

We don’t generate videos. 🎬 We generate worlds from videos. 🌍 Introducing InSpatio-World — the world's first open-source real-time 4D world model‼️ Your input: a video clip Our output: a dynamic, navigable, persistent world 🕹️ explore freely across viewpoints ⏪ control time forward and backward 🔓 open-source and ready to build on :) Live demo: 🔗 Code & weights: 🔗 Project page: 🔗

We don’t generate videos. 🎬 We generate worlds from videos. 🌍 Introducing InSpatio-World — the world's first open-source real-time 4D world model‼️ Your input: a video clip Our output: a dynamic, navigable, persistent world 🕹️ explore freely across viewpoints ⏪ control time forward and backward 🔓 open-source and ready to build on :) Live demo: 🔗 Code & weights: 🔗 Project page: 🔗

InSpatio｜影溯

109,948 次观看 • 5 个月前

Today we’re releasing OpenEQA — the Open-Vocabulary Embodied Question Answering Benchmark. It measures an AI agent’s understanding of physical environments by probing it with open vocabulary questions like “Where did I leave my badge?” More details ➡️ All of today’s state-of-art vision+language models (VLMs) fall well short of human performance. In fact, for questions that require spatial understanding, today’s VLMs are nearly “blind” – access to visual content provides only minor improvements over language-only models. We hope that OpenEQA motivates additional research into helping AI understand and communicate about the world it sees.

Today we’re releasing OpenEQA — the Open-Vocabulary Embodied Question Answering Benchmark. It measures an AI agent’s understanding of physical environments by probing it with open vocabulary questions like “Where did I leave my badge?” More details ➡️ All of today’s state-of-art vision+language models (VLMs) fall well short of human performance. In fact, for questions that require spatial understanding, today’s VLMs are nearly “blind” – access to visual content provides only minor improvements over language-only models. We hope that OpenEQA motivates additional research into helping AI understand and communicate about the world it sees.

AI at Meta

407,109 次观看 • 2 年前

MotionGPT: Human Motion as a Foreign Language paper page: Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.

MotionGPT: Human Motion as a Foreign Language paper page: Though the advancement of pre-trained large language models unfolds, the exploration of building a unified model for language and other multi-modal data, such as motion, remains challenging and untouched so far. Fortunately, human motion displays a semantic coupling akin to human language, often perceived as a form of body language. By fusing language data with large-scale motion models, motion-language pre-training that can enhance the performance of motion-related tasks becomes feasible. Driven by this insight, we propose MotionGPT, a unified, versatile, and user-friendly motion-language model to handle multiple motion-relevant tasks. Specifically, we employ the discrete vector quantization for human motion and transfer 3D motion into motion tokens, similar to the generation process of word tokens. Building upon this "motion vocabulary", we perform language modeling on both motion and text in a unified manner, treating human motion as a specific language. Moreover, inspired by prompt learning, we pre-train MotionGPT with a mixture of motion-language data and fine-tune it on prompt-based question-and-answer tasks. Extensive experiments demonstrate that MotionGPT achieves state-of-the-art performances on multiple motion tasks including text-driven motion generation, motion captioning, motion prediction, and motion in-between.

AK

125,353 次观看 • 3 年前