Animating objects using the human body
Methods of animating objects using the human body are described. In an embodiment, a deformation graph is generated from a mesh which describes the object. Tracked skeleton data is received which is generated from sensor data and the tracked skeleton is then embedded in the graph. Subsequent motion which is captured by the sensor result in motion of the tracked skeleton and this motion is used to define transformations on the deformation graph. The transformations are then applied to the mesh to generate an animation of the object which corresponds to the captured motion. In various examples, the mesh is generated by scanning an object and the deformation graph is generated using orientation-aware sampling such that nodes can be placed close together within the deformation graph where there are sharp corners or other features with high curvature in the object.
1. A method of animating an object comprising:
generating, by a processor, a deformation graph automatically from an input mesh defining the object, the generation comprising:
defining a plurality of node positions using sampling;
creating a nearest neighbor graph by connecting each node to a number of nearest neighbor nodes; and
the sampling and the creating of the nearest neighbor graph both using an orientation-aware distance defined such that distances between a pair of samples increases faster when normals associated with each of the samples deviate from alignment;
receiving body tracking data defining positions of one or more points on a body;
attaching points on a skeleton to points on the deformation graph using the body tracking data;
transforming the deformation graph in real-time based on motion of the tracked skeleton by computing a plurality of transformations; and
dynamically applying the plurality of transformations to the mesh to render a corresponding animation of the object.
2. A method according to claim 1 , further comprising creating the input mesh by:
generating a 3D volumetric reconstruction of a scene scanned by a user with a depth camera;
segmenting an object from the 3D volumetric reconstruction of the scene; and
extracting a geometric isosurface from the segmented portion of the 3D volumetric reconstruction.
3. A method according to claim 1 , wherein attaching points on the skeleton to points on the deformation graph comprises attaching points on the deformation graph to points on the skeleton, the points on the skeleton being between joints of the skeleton.
4. A method according to claim 1 , wherein generating a deformation graph automatically from an input mesh defining the object comprises:
defining a plurality of node positions using sampling;
creating a nearest neighbor graph by connecting each node to a number of nearest neighbor nodes; and
adding additional edges to the nearest neighbor graph based on analysis of connected components in the nearest neighbor graph.
5. A method according to claim 4 , wherein the skeleton is an animal skeleton.
6. A method according to claim 4 , wherein adding additional edges to the nearest neighbor graph based on analysis of connected components in the nearest neighbor graph comprises:
identifying any connected components in the nearest neighbor graph;
forming a weighted graph comprising nodes corresponding to each identified connected component and weights defined based on a minimum distance between the connected components in the nearest neighbor graph;
computing a minimum spanning tree on the weighted graph; and
adding an additional edge to the nearest neighbor graph where there is an edge between the connected components in the spanning tree.
7. A method according to claim 4 , wherein transforming the deformation graph in real-time based on motion of the tracked skeleton by computing a plurality of transformations comprises treating the additional edges as links with greater rigidity than other connections between nodes in the nearest neighbor graph.
8. A method according to claim 1 , wherein attaching points on a skeleton to points on the deformation graph comprises:
providing visual feedback to a user showing alignment of the tracked body and the input mesh;
dynamically generating a proposed mapping between points on a skeleton and a number of nearest nodes in the input mesh: and in response to a verbal command received from the user, storing the proposed mapping,
and wherein the stored mapping is used in dynamically applying the plurality of transformations to the mesh.
9. A method according to claim 1 , wherein the object is non-humanoid.
10. A method according to claim 1 , wherein the method is implemented interactively at run-time.
11. A method as claimed in claim 1 at least partially carried out using hardware logic.
12. A system comprising:
an input for receiving an input mesh defining an object;
a pre-processing module arranged to generate a deformation graph automatically from the input mesh and approximate the input mesh as a series of kinematic sphere proxies;
a warp module arranged to receive the deformation graph, the input mesh and body tracking data defining positions of one or more points on a body, to attach points on a skeleton to points on the deformation graph using the body tracking data, to compute a series of transformations on the deformation graph in real-time based on motion of the tracked skeleton, and to apply the series of transformations to the input mesh to generate an animation of the object; and
at least one of:
the pre-processing module further arranged to define a plurality of node positions using sampling and create a nearest neighbor graph by connecting each node to a number of nearest neighbor nodes, the sampling and the creating of the nearest neighbor graph both using an orientation-aware distance defined such that distances between a pair of samples increases faster when normals associated with each of the samples deviate from alignment;
the warp module further arranged to attach points on a skeleton to points on the deformation graph using the body tracking data, wherein the attachment comprises a plurality of phases wherein a first phase corresponds to at least one of a first set of points on the skeleton or a first set of points on the deformation graph and a second phase corresponds to at least one of a second set of points on the skeleton or a second set of points on the deformation graph; or
the pre-processing module further arranged to approximate the input mesh as a series of kinematic sphere proxies, wherein the approximation comprises placing at least one of the series of kinematic sphere proxies at at least one mesh vertex.
13. A system according to claim 12 , further comprising:
a skeleton tracker module arranged receive data from a sensor and to generate the body tracking data from the received data.
14. A system according to claim 13 , wherein the sensor is a depth camera.
15. A system according to claim 12 , wherein the pre-processing module is arranged to generate the deformation graph using orientation-aware sampling.
16. A system according to claim 15 , wherein the pre-processing module is further arranged to generate the deformation graph by defining a nearest neighbor graph from the input mesh using orientation-aware sampling and then adding additional edges to the nearest neighbor graph based on analysis of connected components in the nearest neighbor graph.
17. A system according to claim 16 , wherein the warp module is arranged to treat the additional edges as more rigid than other edges in the deformation graph when computing the series of transformations on the deformation graph.
18. A system according to claim 12 , further comprising:
a processor; and
a memory arranged to store the pre-processing module and the warp module,
and wherein the pre-processing module and warp module comprise processor-executable instructions.
19. A system as claimed in claim 12 , wherein at least one of the pre-processing module and the warp module being at least partially implemented using hardware logic selected from any one or more of: a field-programmable gate array, a program-specific integrated circuit, a program-specific standard product, a system-on-a-chip, a complex programmable logic device.
20. One or more tangible device-readable media with device-executable instructions comprising:
device-executable instructions to generate a deformation graph automatically from an input mesh defining an inanimate object using orientation-aware sampling;
device-executable instructions to receive body tracking data relating to a user, the data defining positions of one or more points on the user's body;
device-executable instructions to attach points on a skeleton to points on the deformation graph using the body tracking data in response to a user command, the attachment comprising a plurality of phases wherein a first phase corresponds to at least one of a first set of points on the skeleton or a first set of points on the deformation graph and a second phase corresponds to at least one of a second set of points on the skeleton or a second set of points on the deformation graph;
device-executable instructions to transform the deformation graph interactively in real-time based on motion of the tracked skeleton by computing a plurality of transformations, the motion of the tracked skeleton corresponding to motion of the user; and
device-executable instructions to dynamically apply the plurality of transformations to the mesh to render a corresponding animation of the object.