search

nianticlabs / simplerecon

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

Predict Depth Image using multiple RGB images #37

Closed Xuefei98 closed 1 year ago

Xuefei98 commented 1 year ago

First of all! Great work! And thanks for open sourcing it! We are trying to use SimpleRecon to reconstruct the depth image for each frame given a sequence of RGB images that collected by RealSenseD455. I wonder if there is an easy way to feed your custom images into the model? We do not necessary need the whole reconstruction part but the depth prediction. Thank you!

mohammed-amr commented 1 year ago

Hello, thanks!

SimpleRecon is in essence a depth predictor from a stream of posed RGB images. There are multiple examples of datasets for video streams in the repo. Please use those as a reference for writing your own dataset class that inherits from GenericMVSDataset.