Zeinab Sadeghipour

I am an M.Sc. student at the School of Computing Science at Simon Fraser University. My supervisor is Professor Hao (Richard) Zhang, the director of Graphics Union Vision Laboratory. I am mainly intereseted in computer vision, computer graphics, image processing and how these areas can be related.

Before coming to SFU, I completed my undergraduate studies in Computer Engineering department at Sharif University of Technology.

What I seek in my research

Learning 3D Scene Synthesis from Annotated RGB-D Images

My master thesis is focused on developing a data-driven method for synthesizing 3D indoor scenes by inserting objects progressively into an initial, possibly, empty scene. Instead of relying on few hundreds of hand-crafted 3D scenes, we take advantage of existing large-scale annotated RGB-D datasets, in particular, the SUN RGB-D database consisting of 10,000+ depth images of real scenes, to form the prior knowledge for our synthesis task.

Related Publication

Zeinab Sadeghipour Kermani, Zicheng Liao, Ping Tan, and Hao Zhang, "Learning 3D Scene Synthesis from Annotated RGB-D Images", Computer Graphics Forum (Special Issue of SGP), Vol. 35, No. 5, to appear, 2016. [PDF] [bibtex]

3D Mesh Watermarking

In my B.Sc. thesis, I studied and implemented a semi-fragile and blind watermarking algorithm for 3D meshes based on the modification of the Laplacian coordinates and using the Quantization Index Modulation. The scheme is robust against attacks such as translations and rotations due to the invariance of the Laplacian vector lengths under such transformations. Besides, in order to achieve uniform scaling invariance, an adaptive quantization step calculating strategy is employed. The watermark extraction is carried out blindly, with no refrence to the host model.

How I improved my practical skills

Let's take a look at the list of course projects I did during my M.Sc. program:

Mesh Cutting with Semi-automatic Rotation

We present a mesh cutting tool to manually draw the boundary of segmentation while assisting the user with semi-automatic rotation. Based on the current endpoint of the sketch, we predict the potential subsequent position of the boundary point considering the minima rule. After obtaining the complete cutting contour, we segment the input mesh into two separate parts with a region growing algorithm. [PDF]

Removing Shadows from Images

We attempt to recover a 2D chromaticity intrinsic variation of a single RGB image which is independent of lighting, without requiring any prior knowledge about the camera. The proposed algorithm aims to learn an invariant direction for projecting from 2D color space to the greyscale intrinsic image. We notice that along this direction, the entropy in the derived invariant image is minimized. In addition, we examined the idea to utilize projection pursuit instead of entropy minimization to find the desired direction. [PDF]

A Legibility Classifier for Deformed Letters

We introduce an approach for classifying a deformed letter from the English alphabet based on its legibility. Given an image of a deformed letter d, our method proposes the use of the filtered medial axis (i.e. strokes) as its simplified representation, from which we can compute a feature vector. Feature correspondence across deformations is ambiguous; thus we propose using a kernel-based multi-instance classifier model. Our classifier is trained using legibility score labels collected through a user-study. [PDF]

Want to know more? Contact me!

If you are interested in my research and projects or have any questions, feel free to contact me. The easiest way is to email me at zsadeghi [at] sfu [dot] ca.