search

isl-org / MiDaS

Code for robust monocular depth estimation described in "Ranftl et. al., Towards Robust Monocular Depth Estimation: Mixing Datasets for Zero-shot Cross-dataset Transfer, TPAMI 2022"
MIT License
4.27k stars 599 forks source link

Video input flickers badly #222

Open juntaosun opened 1 year ago

juntaosun commented 1 year ago

This is a great project, when I use the video input, the depth picture is flickering badly, will the next version fix it? Thank you so much~

KexianHust commented 10 months ago

@juntaosun You can try our video depth model ViTA is based on DPT (MiDaS 3.0).

Cedvano commented 10 months ago

@juntaosun You can try our video depth model ViTA is based on DPT (MiDaS 3.0).

Not working, error wint torch.load(checkpoint) and cuda.

KexianHust commented 10 months ago

@juntaosun You can try our video depth model ViTA is based on DPT (MiDaS 3.0).

Not working, error wint torch.load(checkpoint) and cuda.

You can change the Line 193 in demo.py:

os.environ["CUDA_VISIBLE_DEVICES"] = "1"

Because our machine has 2 GPUs, so we set 'CUDA_VISIBLE_DEVICES' to 1. If you only have 1 GPU, you can change '1' to '0'.

Cedvano commented 10 months ago

Oh, sorry, I don't see this parameter :(

Now, that's work fine. Thanks.