search

NVIDIA-AI-IOT / trt_pose

Real-time pose estimation accelerated with NVIDIA TensorRT
MIT License
957 stars 290 forks source link

What is cmap_window? Does this need to be changed if you're experimenting with 384*384 networks? #177

Open troymyname opened 2 months ago

troymyname commented 2 months ago

I am having this problem where I am getting key-points, but they are not scaled properly to the image.

For instance, I am deriving key-points using the methods stated. However, the key-point placements are not correct and seems to be a scaling issue.

Does anyone know if this is related to cmap_window parameter in ParseObjects.py.