Out-of-Core Surface Reconstruction via Global TGV Minimization [PDF]
Available on arxiv and presented on ICCV 2021.
We present an out-of-core variational approach for surface reconstruction from a set of aligned depth maps. Input depth maps are supposed to be reconstructed from regular photos or/and can be a representation of terrestrial LIDAR point clouds. Our approach is based on surface reconstruction via total generalized variation minimization (TGV) because of its strong visibility-based noise-filtering properties and GPU-friendliness. Our main contribution is an out-of-core OpenCL-accelerated adaptation of this numerical algorithm which can handle arbitrarily large real-world scenes with scale diversity.
Citywall reconstruction results (better video is encoded with H.265/HEVC):
citywall_reconstruction_h264.mp4
- /pdf/poliarnyi2021.pdf - the paper (compiled pdf)
- /pdf/poliarnyi2021_supp.pdf - the supplementary (compiled pdf)
- /presentation/poliarnyi2021_tgv_ooc_presentation.mp4 - presentation video (with embedded subtitles)
- /presentation/poliarnyi2021_tgv_ooc_slides.pdf - presentation slides (pdf)
- /presentation/poliarnyi2021_tgv_ooc_poster.pdf - poster (pdf)
- /presentation/poliarnyi2021_tgv_ooc_slides_ru.pdf - presentation slides for Yandex Christmas Colloquium on CV 2021 (RU, pdf)
- /anc/citywall_reconstruction.mp4 - supplementary video with Citywall dataset results (H.265/HEVC codec)
- /anc/blocks.txt - supplementary text file with list of the Copenhagen city tiles (for easy automated downloading with Python)
- /ms.tex - the paper source
- /supplement.tex - the supplementary source (with the Copenhagen city results and with minor tables)
- /images/* - images used for the final paper (downscaled to fit in size limits)
- /images_highres/* - original images (including unused) in original high resolution
- How to compile main paper:
pdflatex ms; bibtex ms; pdflatex ms; pdflatex ms; mv ms.pdf poliarnyi2021.pdf
- How to compile supplementary text:
pdflatex supplement; bibtex supplement; pdflatex supplement; pdflatex supplement; mv supplement.pdf poliarnyi2021_supp.pdf
- How to add text to the video:
ffmpeg -i citywall_raw.mp4 -vf "drawtext=text='Citywall dataset':enable='between(t,0,5)': x=(w-text_w)/2: y=(h-text_h)*4/3/2: fontsize=50: fontcolor=black: box=0: [email protected]: boxborderw=5:" -c:a copy output1.mp4
ffmpeg -i output1.mp4 -vf "drawtext=text='Supplementary material for paper\:':enable='between(t,0,5)': x=(w-text_w)/2: y=(h-8*text_h): fontsize=32: fontcolor=black: box=0: [email protected]: boxborderw=5:" -c:a copy output2.mp4
ffmpeg -i output2.mp4 -vf "drawtext=text='Out-of-Core Surface Reconstruction via Global TGV Minimization.':enable='between(t,0,5)': x=(w-text_w)/2: y=(h-7*text_h): fontsize=32: fontcolor=black: box=0: [email protected]: boxborderw=5:" -c:a copy output3.mp4
ffmpeg -i output3.mp4 -vcodec libx265 -crf 28 citywall_reconstruction.mp4
- How to send to arxiv:
4.1) Delete pdf/, axessibility.sty, readme.md, egbib.bib and ieee_fullname.bst
4.2) Pack into zip and upload