
NNFlowVector 是 Foundry Nuke 的光流插件。它使用最先进的人工智能/深度学习技术来产生真正准确的运动矢量。
功能
产生高质量的运动矢量(“前向”和“后向”层共同构成 Nuke 中的“运动”层)。
生成高质量的智能矢量,以便能够将输出与 NukeX 的强大节点(如 VectorDistort、VectorCornerPin 和 GridWarpTracker)一起使用。(此功能需要 NNFlowVector 的许可版本)。
本机过扫描处理(即它处理比图像帧/格式更大的边界框)。请注意,边界框不能设置动画,即它不能因帧而异。
GPU 使用 NVIDIA 的 CUDA 加速(需要 NVIDIA 显卡)。
多个推理/图像补丁的内部拼接(能够处理高分辨率图像以产生运动矢量,而 GPU 上没有那么多 VRAM 可用)。
File size: 2.74 GB
NNFlowVector is an optical flow plugin for Foundry’s Nuke. It’s using state of the art AI/deep learning technology to produce really accurate motion vectors.
Features
Produce high quality motion vectors (“forward” and “backward” layers that together make up the “motion” layer in Nuke).
Produce high quality smart vectors, to be able to use the output with NukeX’s powerful nodes like VectorDistort, VectorCornerPin and GridWarpTracker. (This feature requires a licensed version of NNFlowVector).
Native overscan handling (i.e. it handles larger bounding boxes than the image frame/format). Please note that the bounding boxes can’t be animated, i.e. it can’t be different from frame to frame.
GPU accelerated using CUDA by NVIDIA (requires a NVIDIA graphics card).
Internal stitching of several inference/image patches (to be able to process high resolution images to produce the motion vectors without having that much VRAM available on the GPU).
原创文章,作者:steven,如若转载,请注明出处:https://rr3d.cn/43533.html