Problems with Darknet/YOLO

[oliverguy]

I know this isn't strictly programming but I feel like the folks here should be best equipped to help me with my question.

 

So I have compiled Darknet with CUDA and have tried running 

./darknet detector demo cfg/coco.data cfg/yolo.cfg yolo.weights

 in the terminal.

It shows a window with my webcams output but I don't see any bounding boxes around any objects. If I put in 

-thresh 0.01

 I do see a grid of bounding boxes labled "person". These don't move if I move around the frame, just stay in a fixed position.

If I try just running it over one image - I tired the default dog.jpeg image that is included - I still get not bounding boxes. If I compile without CUDA and OpenCV I get the expected 3 boxes around the dog, car and bike on the default threshold.

 

Any help with sorting this issue would be greatly appreciated - Oliver