Nail Ibrahimli

I am a PhD student at the 3DUU lab and part of the 3D Geoinformation group at Delft University of Technology , where I work on 3D computer vision. My PhD advisor is Liangliang Nan.

I studied BSc in Computer Engineering at Middle East Technical University and gained MSc in Informatics at the Technical University of Munich. I did my master’s thesis project at the Singapore Changi Airport, it was funded by TUMCREATE. The project goal was to provide a simple and portable solution for measuring airfreight shipments.

Before starting my master’s degree I was working on 3D Vision topics at the Italian Institute of Technology Visual Geometry and Modelling Lab. During my masters, I have also spent time working with Stanford Research Institute spinoff REscan Inc. in 3D reconstruction pipeline.

I am mainly interested in 3D understanding and vision-based navigation topics. My doctoral research investigates learning-based 3D reconstruction and 3D urban scene understanding.

Email  /  GitHub  /  LinkedIn

profile photo

Research

I'm interested in 3D reconstruction, vision-based navigation, robotic 3D vision, computer vision, machine learning.

project image

DDL-MVS: Depth Discontinuity Learning for MVS Networks



Nail Ibrahimli, Hugo Ledoux, Julian Kooij, Liangliang Nan
Delft University of Technology.

Traditional MVS methods have good accuracy but struggle with completeness, while recently developed learning-based multi-view stereo (MVS) techniques have improved completeness except accuracy being compromised. We propose depth discontinuity learning for MVS methods, which further improves accuracy while retaining the completeness of the reconstruction. Our idea is to jointly estimate the depth and boundary maps where the boundary maps are explicitly used for further refinement of the depth maps. We validate our idea and demonstrate that our strategies can be easily integrated into the existing learning-based MVS pipeline where the reconstruction depends on high-quality depth map estimation. Extensive experiments on various datasets show that our method improves reconstruction quality compared to baseline. Experiments also demonstrate that the presented model and strategies have good generalization capabilities.

Potential Thesis Projects

List of potential thesis projects, please click on the images for the details. Feel free to contact me if your own thesis topic is aligned with my research focus.

project image

Mesh reconstruction of indoor environments from images



Nail Ibrahimli
Delft University of Technology.

Neural Radiance Fields is a method that achieves state-of-the-art results for synthesizing novel views of complex scenes by optimizing an underlying continuous volumetric scene function using a set of input views.
Multi-View Stereo infers the dense 3D geometry from a set of calibrated image views. It is one of the main components of 3D reconstruction pipelines. Since 2015, deep learning has been increasingly used to solve several 3D vision problems due to its predominating performance, and since 2017 learning-based multi-view stereo problems become a hot topic due to the robustness of CNN to scene variations.
Goal: This project will address the challenge of reconstructing 3D indoor scenes from a set of images. Current photogrammetry approaches have shown accurate and complete reconstruction results on textured objects while struggling with man-made textureless planar environments like man-made spaces. The goal of this project is to incorporate planar constraints into the learning-based 3D reconstruction pipeline where the final output will be complete and accurate mesh.

project image

Multi-view Styled Stereo



Nail Ibrahimli
Delft University of Technology.

Multi-view stereo infers the dense 3D geometry from a set of calibrated image views. It is one of the main components of 3D reconstruction pipelines. Since 2015, deep learning has been increasingly used to solve several 3D vision problems due to its predominating performance, and since 2017 learning-based multi-view stereo problems become a hot topic due to the robustness of CNN to scene variations.
Neural Style Transfer In fine art, especially painting, humans have mastered the skill to create unique visual experiences through composing a complex interplay between the content and style of an image. There are Deep Learning methods that are using neural representations to composite content and style of arbitrary images, providing a neural algorithm for the creation of artistic images.
Goal: The goal of this project is styled dense 3D reconstruction from a visual set of content images and a style image.

project image

Multi-view Semantic Stereo



Nail Ibrahimli
Delft University of Technology.

Multi-view stereo infers the dense 3D geometry from a set of calibrated image views. It is one of the main components of 3D reconstruction pipelines. Since 2015, deep learning has been increasingly used to solve several 3D vision problems due to its predominating performance, and since 2017 learning-based multi-view stereo problems become a hot topic due to the robustness of CNN to scene variations.
Semantic segmentation is the task of clustering parts of an image/pointcloud together which belong to the same object class. It is a form of pixel-level/point-level prediction because each pixel/point in an image/pointcloud is classified according to a category.
Goal: The goal of this project is semantically aware 3D reconstruction from a visual set of content images.

Teaching

project image

Machine Learning for the Built Environment



Liangliang Nan, Shenglan Du, Nail Ibrahimli
Delft University of Technology.
2021-2022 Q3

This course is introductory for machine learning to equip students with the basic knowledge and skills for further study and research in machine learning. It introduces the theory/methods of well-established machine learning and a few state-of-the-art deep learning techniques for processing geospatial data (e.g., point clouds). Students will also gain hands-on experiences by applying commonly used machine learning techniques to solve practical problems through a series of lab exercises and assignments.

project image

Photogrammetry and 3D Computer Vision



Liangliang Nan, Nail Ibrahimli
Delft University of Technology.
2020-2021 Q3, 2021-2022 Q4

Photogrammetry and 3D Computer Vision (i.e., 3DV) aim at recovering the structure of real-world objects/scenes from images. This course is about the theories, methodologies, and techniques of 3D computer vision for the built environment. In the term of this course, you will learn the basic knowledge and algorithms in 3D computer vision through a series of lectures, reading materials, lab exercises, and group assignments.





Design and source code from Jon Barron's website and Leonid Keselman 's Jekyll fork
Profile photo from Nadine Hobeika