what's the preprocess and postprocess pipeline of RTMDet and RTMPose?

[RachelLYY]

I tried using onnxruntime to inference rtmdet and rtmopse.
I can run the code following the guide code, but the output seems wrong.
The preposs of rtmdet is: padding and resize img to (1,3, 320,320). The postprocess of rtmdet is: nms.
The preposs of rtmdet is: padding and resize img to (1,3,256,192)
For rtmdet, the output shape is (1,100,5), but the output has losts of numbers less than zero.

For rtmpose, the output shape is (1,17,384) and the output each number is less than 1.

I think this should be because I didn't do the data pre and pose processing right. Could you help me with it? Thanks!