text stringlengths 0 596 |
|---|
For example, if a wheel rotates clockwise in the reference but the corresponding wheel rotates counterclockwise in the generated result, this should not be penalized if the generated mechanism still realizes mechanically equivalent component relationships, interactions, and functional outcomes. |
A difference in motion direction should be considered an error only when it breaks the mechanical relationship, causes an incorrect interaction, or changes the intended mechanical function. |
### C. Relative Motion and Kinematic Consistency |
* Do connected components exhibit reasonable relative motion? |
* Do hinged components rotate around appropriate connection points? |
* Do components with fixed mechanical relationships maintain the expected relative constraints? |
* Are the motion relationships among gears, rods, shafts, tracks, and other components mechanically reasonable? |
* Do components exhibit clearly disconnected, drifting, or independent motion inconsistent with their mechanical relationships? |
If such problems emerge during dynamic execution, they should lower the Mechanical Motion score. |
### D. Dynamic Connectivity and Constraint Consistency |
Evaluate whether mechanical connections remain valid throughout the entire motion process. |
Consider: |
* Do components that are initially connected remain connected during motion? |
* Do connections between rods and gears, rods and shafts, linkages and hinges, or other connected components remain correct throughout execution? |
* Do connected components gradually separate, detach, drift, or become misaligned during motion? |
* Do hinge or joint locations move incorrectly, causing components that should remain connected to separate? |
* Are mechanical constraints correct in some frames but broken in later frames? |
If a rod is correctly connected to a gear initially but gradually separates from the gear during motion, this is a dynamic connectivity failure and should lower the Mechanical Motion score. |
If the rod is already incorrectly disconnected from the gear in the initial static configuration, the connectivity error should lower the Visual Structure score. If this error also causes the actual mechanical execution to fail, the Mechanical Motion score should also be lowered. |
### E. Mechanical Interaction |
* Do important contacts, collisions, pushes, transfers, engagements, or other mechanical interactions occur correctly? |
* Does the motion of one component form a reasonable mechanical interaction with the response of subsequent components? |
* Are important interactions clearly demonstrated in the reference missing from the generated result? |
* Are there clearly incorrect contacts, collisions, or motion transfers? |
* In multi-stage mechanisms, do the different stages form a coherent sequence of mechanical interactions? |
If a static structural error causes an intended mechanical interaction to fail during execution, the Mechanical Motion score should be lowered. |
### F. Dynamic Geometric Consistency |
Consider: |
* Do components maintain reasonable spatial relationships during motion? |
* Does any generated frame show obvious interpenetration or components passing through each other? |
* Are there nonphysical overlaps, abnormal crossings, or severe geometric conflicts? |
* Do components suddenly detach from the mechanical structure? |
* Do components jump to spatially discontinuous positions? |
* Is there obvious abnormal deformation or spatial discontinuity? |
Even if most frames appear correct, obvious dynamic penetration, separation, or other severe mechanical failures in some frames must be reflected in the Mechanical Motion score. |
If two components are already incorrectly intersecting in the initial static configuration, the Visual Structure score should be lowered. |
If the initial configuration is valid but interpenetration occurs during motion, the Mechanical Motion score should be lowered. |
If a static geometric error further causes actual dynamic execution failure, both Visual Structure and Mechanical Motion may be penalized accordingly. |
### G. Temporal Progression |
* Does the generated sequence demonstrate a coherent mechanical process over time? |
* Are the temporal relationships between different motion stages mechanically reasonable? |
* Do necessary preceding interactions occur before the corresponding downstream behavior? |
* Are there obvious ordering errors or discontinuous changes? |
* Do components suddenly change state without a reasonable mechanical progression being supported by the generated sequence? |
The reference sequence and generated sequence are two independent temporal sequences with no correspondence in timestamps or frame positions. |
Therefore, do not directly compare the i-th frame of one sequence with the i-th frame of the other, and do not judge correctness based on where an event occurs within the relative frame positions of the two sequences. |
Differences in animation speed should not be considered errors. |
### H. Motion Completeness |
* Does the generated animation cover the important mechanical stages shown in the reference? |
* Are any critical motion stages, mechanical interactions, or final outcomes missing? |
* Does the animation stop before the core mechanical process is completed? |
* Does the complete mechanical system progress from a reasonable initial state to the intended final state or outcome? |
For cyclic mechanisms: |
* Determine whether the generated animation demonstrates at least one complete motion cycle or sufficiently covers all key stages of the cycle. |
For non-cyclic mechanisms: |
* Determine whether the generated animation progresses through all important mechanical stages and completes the intended mechanical process. |
Because only temporally sampled frames are provided, a brief intermediate event does not need to be directly captured in a sampled frame. If other frames in the generated sequence reasonably support that a brief event occurred, do not penalize the result merely because that event was not directly sampled. |
### I. Motion Observability |
The generated result must sufficiently present its mechanical motion through its own camera and rendered frames. |
Consider: |
* Are the key moving components visible in the generated frames? |
* Can important mechanical interactions be observed? |
* Does the camera continue to show the important regions of the mechanism during execution? |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.