We are currently working on the detection of ice floes through a stereo camera mounted to the side of an ice breaker. We are using a ZED module and with the images we have managed to create a roboflow model that seems very functional.
Currently we are on the latest version of unbuntu, python 3.10.11 and CUDA 11.7. The problem we encounter comes when trying to apply the model on a video directly on our system through the infer.sh program available on
after this input
ROBOFLOW_KEY=“Our_Key” ./infer.sh [Our_Project]/[3] Video.mp4 Video_1.mp4
in the python terminal
we receive the same message regardless of which format the videos are set to (mp4, mov, avi)
The videos we are using are currently five minutes long and about 15mo. Is there any advice you guys can give us