Im writing a c++ application that already uses succesfully many of hailo official post processing scripts for other models like yolo object detection, scrfd, etc. Today i was trying to make yolov8 pose estimation work but got stuck on this post-process (yolov8pose_postprocess.cpp) issue: the model runs fine, no crashes, etc. however the final output is wrong and out of the [0.0,1.0] bounds:
{-0.577835 , -0.877895 , 3.39016e-05}
{-0.579439 , -0.878667 , 3.5257e-05}
{-0.57831 , -0.877538 , 3.40429e-05}
{-0.577598 , -0.878607 , 3.32042e-05}
{-0.57831 , -0.879558 , 3.40227e-05}
{-0.575044 , -0.878489 , 3.3581e-05}
{-0.577716 , -0.877241 , 3.3761e-05}
{-0.572549 , -0.878726 , 3.65578e-05}
{-0.577776 , -0.881042 , 3.3761e-05}
{-0.573381 , -0.878904 , 3.28901e-05}
{-0.578073 , -0.876528 , 3.36409e-05}
{-0.576766 , -0.878845 , 3.53408e-05}
{-0.578132 , -0.881577 , 3.36209e-05}
{-0.568451 , -0.878548 , 3.51107e-05}
{-0.578429 , -0.870708 , 3.36409e-05}
{-0.568748 , -0.878667 , 3.53618e-05}
{-0.578251 , -0.877063 , 3.36209e-05}
can somebody help?
I have tried using both yolov8s-pose and yolov8m-pose as well as both both commit versions of
tappas/core/hailo/libs/postprocesses/pose_estimation
/yolov8pose_postprocess.cpp
without any luck. can somebody help? im using hailoRT 4.18 and hailo-8L on rpi