Click here for the source code of the CoSPAIR descriptor proposed in our following publication:
B.Logoglu, S.Kalkan, A.Temizel, “CoSPAIR: Colored Histograms of Spatial Concentric Surflet-Pairs for 3D Object Recognition”, Robotics and Autonomous Systems, vol. 75, part B, pp. 558-570, doi: 10.1016/j.robot.2015.09.027, Jan. 2016.
I am looking for MSc/PhD students in METU to work on GPU development in a TÜBİTAK funded project to work in collaboration with TÜBİTAK-UZAY. The scholarship will start in June 2017 and will be for a fixed term of 36 months. Please send me an e-mail with your CV if you are interested.
I’ll be giving a seminar on Multimodal Egocentric Activity Recognition at the Computer Science Department – University of Bristol on 7th Feb 2017 at 12:00.
In the talk, I will be focussing on the particular challenges regarding the activity recognition from egocentric (first-person) data and data-driven selection and weighting of different features.
You can access the slides for Intermediate-Advanced CUDA seminar (Dec 2014) here
The general overview of the seminar is as follows:
1. Introduction and Recent Developments on GPU and CUDA
2. CUDA Performance Considerations Review
3. Background Subtraction Case Study
4. CUDA Profiler and Parallel Nsight
5. Parallel Reduction
6. Thrust Library
7. Multi GPU Programming
8. Advanced Concepts
• Dynamic Parallelism
• HyperQ
• WARP Operations and Shuffling
• CUDA 6.0 Unified Memory
Congratulations to Serdar Buyuksarac who won the “ISRA VISION AG – Computer Vision Award” (1st place) with his master’s thesis ‘GPU-Enabled Real-Time Panoramic Background Subtraction’ that I co-supervised together with his main supervisor Prof.Dr. Gozde Akar.
I’ll be taking sabbatical leave as visiting academic at University of Birmingham between May 2016-May 2017. You can still contact me through my METU e-mail.
Congratulations to Ayse Elvan Gunduz who won the “ISRA VISION AG – Computer Vision Award” (2nd place) with her master’s thesis ‘Anomaly Detection using Sparse Features and Spatio-Temporal Hidden Markov Model for Pedestrian Zone Video Surveillance’ that I co-supervised together with her main supervisor Asst.Prof.Dr. Tuğba Taşkaya Temizel.
Click here for the source code of the GPU based Fast Omnidirectional Image Unwrapping Algorithm described in our publication:
E. Kepucka, I. Gurcan, A. Temizel, “Fast Omnidirectional Image Unwrapping on GPU”, Euromicro International Conference on Parallel, Distributed and Network-Based Computing (WIP), Feb 2012.
We have released our CPU-GPU Hybrid implementation of TLD algorithm which provides 10.25 times speed-up at 1920×1080 resolution compared to the baseline TLD.
Click here for the Open Source Hybrid CPU-GPU Tracking-Learning-Detection (H-TLD) Algorithm