When I read the code, I find the motion-length in test stage is obtained via the length = len(motion) in Pipeline Module: Crop (same to train stage). However, in test stage, we can only input 'Text', and doing it like above is not reasonable. From the paper, we know that the length of motion to retrieval is critical, and I'm wondering if there's something I'm overlooking?
Hi, mingyuan, thanks for your excellent work.
When I read the code, I find the motion-length in test stage is obtained via the
length = len(motion)
inPipeline Module: Crop
(same to train stage). However, in test stage, we can only input 'Text', and doing it like above is not reasonable. From the paper, we know that the length of motion to retrieval is critical, and I'm wondering if there's something I'm overlooking?Looking forward your reply!