search

PengfeiRen96 / DIR

[ICCV 2023 Oral] Decoupled Iterative Refinement Framework for Interacting Hands Reconstruction from a Single RGB Image
https://pengfeiren96.github.io/DIR/
MIT License
69 stars 2 forks source link

Coordinate problem #6

Open luckyday2022 opened 10 months ago

luckyday2022 commented 10 months ago

When training a model, don’t we need to subtract the root joint from the model’s output (out)? 1701700342013 1701700423190

Why are GT and OUT inconsistent?

walsvid commented 9 months ago

Hi @luckyday2022 , the difference between the two lies in the definition. For the position in the gt data set to be an absolute position, the center of gt needs to be subtracted. For network prediction, our definition is the position offset relative to the center. So there is no need to subtract center again when outputting.