Open bestsunlin opened 1 year ago
Hi, thank you for taking the time to try out our model! Can you share a code snippet showing how you are running ZeroDepth? What sort of GPU are you using?
with torch.no_grad():
I use the ZeroDepth on nuScenes images 900x1600 with batch_size=1 and use a single A6000. It takes 5-6 second for one image. It takes more times on decoder than encoder. I see that the image encoder use a lightweight R18. So what causes the low running speed?
When I run your ZeroDepth model, it took very long (seems forever) to run one example with the resolution 768x1024. When I changed the resolution to 256x256, it also takes about 5 mins. Is it because of the model complexity or others?