ADEPT: advanced dense prediction

Semantic analysis of natural images at the pixel level

Reconstruction workshop


Deep learning is most easily applied to recognition tasks. Hence, reconstruction approaches are somewhat less considered these days. This is unfortunate since deep learning can contribute significant improvement to stereo, structure and motion, sparse matching and other reconstruction tasks. This workshop attempts to correct this by encouraging sharing of knowledge and exchange of ideas in this fascinating field.

When and Where

The workshop will be held on Friday, December 17 2021 at 13h in D305. Remote participation is possible over the following teleconference link.


13:10 Ivan Bilić
Rethinking ego-motion for self-supervised monocular depth estimation
Primjena geometrije više pogleda u dubokim modelima za rekonstrukciju pomaka kamere
13:30 Marin Oršić
Leveraging attention for sparse matching
Rijetko uparivanje primjenom slojeva pažnje
13:50 Iva Sović
Self-supervised learning of stereroscopic reconstruction
Samonadzirano ucenje stereoskopske rekonstrukcije
14:10 Pauza
14:20 Filip Oreč
Rimac automobili
Calibration of relative pose between RGB camera and LIDAR
Kalibracija međusobnog položaja kamere i LIDAR-a
14:40 Antonio Jurić
Gideon Brothers
Deep Learning in Visual SLAM
Duboko učenje za vizualnu navigaciju i kartiranje
15:00 Kristijan Bartol
Generalizable Human Pose Triangulation
Trijangulacija položaja osobe