每条样本提供：RGB 源视频、XDoG 线稿视频（生成目标）、目标实例掩码（mask 版）、 SpaTracker 运动轨迹、首/末帧图、自然语言描述。 - 仓库文件：dataset1.zip（21.9 GB），解压后即本文档描述的 datasets/ 目录 （3 个 mask 版子集 + 3 个 dense 版子集，共 2,258 条索引记录） - 解压：unzip dataset1.zip -d SketchTraj/（解压后约 25 GB） 轨迹有两个版本，并存于 datasets/ 下： - mask 版（datasets/ /）：查询点限制在 SAM2 实例掩码内，聚焦主体运动 - dense 版（datasets/ dense/）：全图均匀网格 1024 点，覆盖整帧运动场 只做轨迹与可视化，视频 / 线稿 / 首末帧 / prompt 全部复用 mask 版（不复制文件， 索引里的 sourcevideo / gtvideo 仍指向 datasets/ /）： - trajvis/.mp4 是轨迹叠在 RGB 源视频上的可视化（H.264 / 832×480 / 49 帧 / 12fps）， trajvis/trajreport.json 给出每条的 visibleratio、meansteppx（mask 版还有 inmaskratio） 训练特征（preprocessed/：Wan2.1 VAE latent / UMT5-XXL / OpenCLIP / ycond）当前未产出， 数据集定稿后由 bash…
Independent publisher
Han
Randyhan
Models in Library0
Datasets in Library1
Models on Hugging Face—
Followers—
Compare the published evidence
This profile connects the 1 model and dataset record shown below to their upstream cards and sources. The library counts describe SAVRN's coverage, rather than this publisher's entire output.
Published tasks: Video to video.
Reported licenses: other. Open a record to review its specific license and conditions.
Compare model specifications, source dates, and hardware estimates on the individual records. Publisher identity verification is separate from a model's quality, deployment readiness, or suitability for your work.