Advanced Deep Learnin ing for Computer Vis isio ion Prof. - PowerPoint PPT Presentation

Advanced Deep Learnin ing for Computer Vis isio ion Prof. Leal-Taixé and Prof. Niessner 1

The Team Lecturers Prof. Dr. Laura Prof. Dr. Matthias Leal-Taixé Niessner Tutors Tim Dave Ji Maxim Chen Hou Meinhardt Maximov Prof. Leal-Taixé and Prof. Niessner 2

What is is this is cours rse about • Presentation of advanced Deep Learning methods for various Computer Vision tasks • Focus on new methods, some of them presented only this year! There will be extra references, many opportunities for you to dig deeper into the topics • Research-oriented course Prof. Leal-Taixé and Prof. Niessner 3

While we go over new methods… • You have to come up with your own ideas to solve a specific vision problem! • Strong focus on the practical side: semester-long project where you can put all the knowledge to practice Prof. Leal-Taixé and Prof. Niessner 4

Course o organiz izatio ion Prof. Leal-Taixé and Prof. Niessner 5

About the le lecture Theory: 12 lectures • Every Monday 10 10:0 :00-11 11:3 :30h h • Seminar Room, 02.13.010 • Practical: • • Project to be done in groups of 2 (non-negotiable!) • Presentations during the semester • Wednesdays 14:0 :00-15 15:3 :30h h (Seminar Room, 02.09.023) • Final poster presentation https://dvl.in.tum.de/teaching/adl4cv-ws19/ Prof. Leal-Taixé and Prof. Niessner 6

Gra radin ing system th Febru Exam: 27 27 th ruary, 13 13:3 :30-14 14:30 • Review: 2 review sessions • Practical part = 2/3 of the grade • Exam = 1/3 of the grade • https://dvl.in.tum.de/teaching/adl4cv-ws19/ Prof. Leal-Taixé and Prof. Niessner 7

Pro roje ject deadline 21.10., today: project presentation • 23.10.: project assignments (projects <-> TAs) • 30.10 .10., mid idnig ight: delive liver r a 1 1 page abstract of f your r id idea • fo for r th the pro roje ject. Until 6.11.: Evaluation of the project and feedback • Prof. Leal-Taixé and Prof. Niessner 8

Pro roje ject evaluation Presentations: everyone needs to attend! • Firs irst t pre resentatio ion: firs first re result lts, challe llenges • 04 04.12 12.: Gro roups #1 1 – 11 11.12 .12.: : Gro roups #2 – Prof. Leal-Taixé and Prof. Niessner 9

Pro roje ject evaluation Presentations: everyone needs to attend! • Second pre resentation: alm lmost t fin final l re result lts, new th thin ings • you trie tried 08 08.01. 1.: Gro roups #1 1 – 15 15.0 .01. 1.: : Gro roups #2 – Prof. Leal-Taixé and Prof. Niessner 10

Pro roje ject evaluation Presentations: everyone needs to attend! • 04.0 .02.: .: fin final l deadlin line on re report (d (deadlin line noon) • Max 4 pages using CVPR template – Fin inal l pre resentation = POSTER • – Date 05.02. 13:00-16:00 Prof. Leal-Taixé and Prof. Niessner 11

Gra radin ing system Exam = 1/3 of the grade • Practical part = 2/3 of the grade • Presentations (2 oral pres. + 1 poster) = 1/3 – Final report = 1/3 – Code/submission = 1/3 – Prof. Leal-Taixé and Prof. Niessner 12

Foll llowin ing up wit ith the pro rojects Each project will be assigned to a TA and you will • have weekly office hours to discuss the progress These will be announced after the projects are • approved Prof. Leal-Taixé and Prof. Niessner 13

Sli lides • Moodle is set up! Lecture will NOT be recorded. • Slides will be posted on Moodle and on the website: https://dvl.in.tum.de/teaching/adl4cv-ws19/ • Questions regarding organization of the course: adl4cv@dvl.in.tum.de • Emails to our individual addresses will not be answered! Prof. Leal-Taixé and Prof. Niessner 14

Teams • Teams of two per project! • Moodle is set up! • If you do not have a team – Chat after the lecture – Post it on Moodle Prof. Leal-Taixé and Prof. Niessner 15

Proje ject Id Ideas / Dir irectio ions Prof. Leal-Taixé and Prof. Niessner 16

3D Scene Unders rstandin ing Ji Ji Hou Prof. Leal-Taixé and Prof. Niessner 17

Pro roje jects Dir irections 3D Detection/Segemntation/In Instance/Comple letion on on various 3D data Prof. Leal-Taixé and Prof. Niessner 18

Pro roje ject Dir irections • 3D Detectio ion on Sin Single RGB-D Image. – Song, Shuran, and Jianxiong Xiao. "Deep sliding shapes for amodal 3d object detection in rgb-d images." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2016. – Qi, Charles R., et al. "Frustum pointnets for 3d object detection from rgb-d data." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2018. – Qi, Charles R., et al. "Deep Hough Voting for 3D Object Detection in Point Clouds." arXiv preprint arXiv:1904.09664 (2019). Prof. Leal-Taixé and Prof. Niessner 19

Pro roje ject Dir irections • Lift iftin ing 2D 2D det etectio ion to to 3D – Srivastava, Siddharth, Frederic Jurie, and Gaurav Sharma. "Learning 2D to 3D Lifting for Object Detection in 3D for Autonomous Vehicles." arXiv preprint arXiv:1904.08494(2019). – Kulkarni, Nilesh, et al. "3D-RelNet: Joint Object and Relational Network for 3D Prediction." arXiv preprint arXiv:1906.02729 (2019). – http://www.cvlibs.net/datasets/kitti/ Prof. Leal-Taixé and Prof. Niessner 20

Pro roje ject Dir irections • Inst stance Se Segmenta tatio ion/Completio ion on 3D re reconstr tructio ion – Hou, Ji, Angela Dai, and Matthias Nießner. "3d-sis: 3d semantic instance segmentation of rgb-d scans." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2019. – Hou, Ji, Angela Dai, and Matthias Nießner. "3D-SIC: 3D Semantic Instance Completion for RGB-D Scans." arXiv preprint arXiv:1904.12012 (2019). Prof. Leal-Taixé and Prof. Niessner 21

Pro roje ject Dir irections • 3D Det etectio ion on on Multi-Vie iews – Chen, Xiaozhi, et al. "Multi-view 3d object detection network for autonomous driving." Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2017. – Single View + Merging Prof. Leal-Taixé and Prof. Niessner 22

Pro roje ject Dir irections • How to to co combine ge geometry ry and co color r (a (and ra radar) r) • Dai, Angela, and Matthias Nießner. "3dmv: Joint 3d-multi-view prediction for 3d semantic scene segmentation." Proceedings of the European Conference on Computer Vision (ECCV) . 2018. • Jaritz, Maximilian, Jiayuan Gu, and Hao Su. "Multi-view PointNet for 3D Scene Understanding." arXiv preprint arXiv:1909.13603 (2019). Prof. Leal-Taixé and Prof. Niessner 23

Pro roje ject Dir irections • 3D Reconstru ructi tion fro from RGB im image(s) Choy, Christopher B., et al. "3d-r2n2: A unified approach for single • and multi-view 3d object reconstruction." European conference on computer vision . Springer, Cham, 2016. Fan, Haoqiang, Hao Su, and Leonidas J. Guibas. "A point set • generation network for 3d object reconstruction from a single image." Proceedings of the IEEE conference on computer vision and pattern recognition . 2017. Prof. Leal-Taixé and Prof. Niessner 24

3D vis isio ion and NLP Dave Z. . Chen Prof. Leal-Taixé and Prof. Niessner 25

Pro roje ject Dir irections • 3D D Cross-modal Retri trieval: : Brid ridging the the Gap be betw tween 3D Obje bjects an and Natu atural l La Language De Desc scriptio tions Chen et al. "Text2Shape: Generating Shapes from Natural Language by Learning Joint – Embeddings" ArXiv Preprint. 2018. – Han et al. "Y2Seq2Seq: Cross-Modal Representation Learning for 3D Shape and Text by Joint Reconstruction and Prediction of View and Word Sequences" The AAAI Conference on Artificial Intelligence. 2018. Tutor: Dave Z. Chen – – Contact: zhenyu.chen@tum.de Prof. Leal-Taixé and Prof. Niessner 26

Pro roje ject Dir irections • Auto tomatic ic Descrip ription Genera rating fo for r 3D CAD models ls – Xu et al. "Show, Attend and Tell: Neural Image Caption Generation with Visual Attention" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2016. Lu et al. "Knowing When to Look: Adaptive Attention via A Visual Sentinel for Image – Captioning" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2017. – Tutor: Dave Z. Chen – Contact: zhenyu.chen@tum.de Prof. Leal-Taixé and Prof. Niessner 27

Pro roje ject Dir irections • Scan2Cap: : Genera rating descrip riptions fo for r objects in in 3D scenes – Xu et al. "Show, Attend and Tell: Neural Image Caption Generation with Visual Attention" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2016. Lu et al. "Knowing When to Look: Adaptive Attention via A Visual Sentinel for Image – Captioning" Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition . 2017. – Tutor: Dave Z. Chen – Contact: zhenyu.chen@tum.de Prof. Leal-Taixé and Prof. Niessner 28