Create a segmentation mask with vos

wenbowen123 / BundleTrack

[IROS 2021] BundleTrack: 6D Pose Tracking for Novel Objects without Instance or Category-Level 3D Models

Other

610 stars 66 forks source link

Create a segmentation mask with vos #17

Closed bgyooPtr closed 2 years ago

bgyooPtr commented 2 years ago

I am trying to generate a mask through run_video.py of transductive-vos.pytorch . However, after a few iterations, a CUDA out of memory error occurs. How do I solve it?

My GPU is nvidia gtx titan and it has about 24GB of memory. Error message: CUDA out of memory. Tried to allocate 6.95 GiB (GPU 0; 23.65 GiB total capacity; 9.53 GiB already allocated; 4.29 GiB free; 17.11 GiB reserved in total by PyTorch)

bgyooPtr commented 2 years ago

Resolved by changing the image resolution.