Topics for this year's seminar include computational photography, deep-learning-based 2D image processing and 3D shape learning which can bedefined as follows:

'Computational photography extends digital photography by providing the capability to record much more information and by offering the possibility of processing this information afterward'. (Oliver Bimber)

With the rapid development of Machine Learning methods, especially neural networks, more tasks concerning computational photography can be better handled and thus may yield some quite interesting applications, e.g., image completion, HDR image reconstruction, color segmentation and image colorization, or even image style transfer. 

Additionally, it is also possible to infer 3D shape information from 2D images. Learning 3D models from images or even sketches provides a powerful paradigm for geomertic modeling. Features of corresponding 3D models may also be learned with new learning algorithms.

Going beyond Euclidean data, Deep Learning has also been used to deal with tasks concerning geometry or graphs, which is a relatively new yet promising research area.

Die Erfolgskontrolle erfolgt durch Bewertung der Präsentation (70%) und der Ausarbeitung des Vortragsmanuskriptes (30%).

Seats are limited in this seminar, students who are interested in attending this seminar please send a contact email with your name and student number to this email address: chengzhi.wu@kit.edu

First come, first served. ;-)

Themenliste

Below are the relevant topics and possible papers that will be discussed in this seminar.

• 2D image processing
  • Image quilting
    • Efros et. al., "Image Quilting for Texture Synthesis and Transfer"
  • PatchMatch
    • Barnes et. al., "PatchMatch: A Randomized Correspondence Algorithm for Structural Image Editing"
  • Image colorization
    • Zhang et. al., "Real-Time User-Guided Image Colorization with Learned Deep Priors"
  • Image melding
    • Darabi et. al., "Image Melding: Combining Inconsistent Images using Patch-based Synthesis"
  • Image completion
    • Lizuka et. al., "Globally and Locally Consistent Image Completion"
  • Style transfer
    • Isola et. al., "Image-to-Image Translation with Conditional Adversarial Networks"
    • Liao et. al., "Visual Attribute Transfer through Deep Image Analogy"
    • Zhu et. al., "Unpaired Image-to-Image Translation using Cycle-Consistent Adversarial Networks"
• from 2D to 3D
  • ​Learning 3D shapes from 2D images
    • Kanazawa et. al., "Learning Category-Specific Mesh Reconstruction from Image Collections"
    • Zuffi et. al, "Lions and Tigers and Bears: Capturing Non-Rigid, 3D, Articulated Shape from Images "
  • Texture learning
    • Saito et. al., "Photorealistic Facial Texture Inference Using Deep Neural Networks"
    • Berkiten et. al., "Learning Detail Transfer based on Geometric Features"
• 3D data processing
  • ​Octree-based
    • Riegler et.al., "OctNet: Learning Deep 3D Representations at High Resolutions"
    • Wang et. al., "O-CNN: Octree-based Convolutional Neural Networks for 3D Shape Analysis"
    • Qi et. al., "Volumetric and Multi-View CNNs for Object Classification on 3D Data"
  • Point cloud-based
    • Qi et. al, "PointNet: Deep Learning on Point Sets for 3D Classification and Segmentation"
    • Qi et. al., "PointNet++: Deep Hierarchical Feature Learning on Point Sets in a Metric Space"
    • Guerrero et. al., "PCPNET Learning Local Shape Properties from Raw Point Clouds"
  • Tangent convolution
    • tatarchenko et. al., "Tangent Convolutions for Dense Prediction in 3D"