search

nianticlabs / simplerecon

[ECCV 2022] SimpleRecon: 3D Reconstruction Without 3D Convolutions
Other
1.28k stars 120 forks source link

Evaluation on DTU and Tanks&Temples #33

Closed eugenelyj closed 1 year ago

eugenelyj commented 1 year ago

Hi, i am curious that if the proposed method can work on popular benchmarks that are used in MVS, like DTU and Tanks&Temples?

mohammed-amr commented 1 year ago

SimpleRecon is for monocular depth estimation from posed video. We do show how metadata can help improve scores for a general MVS method, CasMVSNet, on DTU. See S8 in the supplemental of the paper.

UsamaHasan commented 1 year ago

@eugenelyj any luck evaluating it on the DTU dataset?

eugenelyj commented 1 year ago

@UsamaHasan No, i didn't try...

UsamaHasan commented 1 year ago

snapshot01 I did try but the results are pretty strange on tanks and temples, the same goes for DTU scans.

CLinvx commented 1 month ago

snapshot01 I did try but the results are pretty strange on tanks and temples, the same goes for DTU scans.

Hi, have you save this problem? ^_^