ACAkin Caliskan

Akin Caliskan

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article


Akin Caliskan has a total of 19 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • MVL-Net: Pairwise Learning for Multi-View Multiple People Labelling

    by Yue Zhang, Akin Caliskan, Mai Xu, Adrian Hilton, Jean-Yves Guillemaut

    In the multi-view domain, it is challenging to correctly label multiple people across viewpoints because of occlusions, visual ambiguities, appearance variation, etc. Deep learning, although having witnessed remarkable success in computer vision tasks, still remains underexplored for the multi-view labelling task, due to the lack of labelled multi-view datasets. In this paper, we propose a novel end-to-end deep neural network named Multi-View Labelling network (MVL-net) that addresses this issue. To overcome the dataset shortage, a large-scale multi-view dataset is generated by combining 3D human models and panoramic backgrounds, along with human poses and realistic rendering. In the proposed MVL-net, we first incorporate Transformer blocks to capture the non-local information for multi-view feature extraction. A matching net is then introduced to achieve multiple people labelling, by predicting matching confidence scores for pairwise instances from two views, thus addressing the problem of the unknown number of people when labelling across views. An additional geometry feature obtained from the epipolar geometry is integrated to leverage multi-view cues during training. To the best of our knowledge, the MVL-net is the first work using deep learning to train a multi-view labelling network. Comprehensive experiments on both synthetic and real-world datasets demonstrate the effectiveness of the proposed method, which outperforms the existing state-of-the-art approaches.


  • PAV: Personalized Head Avatar from Unstructured Video Collection

    by Akin Caliskan, Berkay Kicanaoglu, Hyeongwoo Kim


  • Open Access

    Learning Dense Wide Baseline Stereo Matching for People

    by Akin Caliskan, Armin Mustafa, Evren Imre, Adrian Hilton

    Existing methods for stereo work on narrow baseline image pairs giving limited performance between wide baseline views. This paper proposes a framework to learn and estimate dense stereo for people from wide baseline image pairs. A synthetic people stereo patch dataset (S2P2) is introduced to learn wide baseline dense stereo matching for people. The proposed framework not only learns human specific features from synthetic data but also exploits pooling layer and data augmentation to adapt to real data. The network learns from the human specific stereo patches from the proposed dataset for wide-baseline stereo estimation. In addition to patch match learning, a stereo constraint is introduced in the framework to solve wide baseline stereo reconstruction of humans. Quantitative and qualitative performance evaluation against state-of-the-art methods of proposed method demonstrates improved wide baseline stereo reconstruction on challenging datasets. We show that it is possible to learn stereo matching from synthetic people dataset and improve performance on real datasets for stereo reconstruction of people from narrow and wide baseline stereo data.

  • Learning Dense Wide Baseline Stereo Matching for People

    by Akin Caliskan, Armin Mustafa, Evren Imre, Adrian Hilton

    Existing methods for stereo work on narrow baseline image pairs giving limited performance between wide baseline views. This paper proposes a framework to learn and estimate dense stereo for people from wide baseline image pairs. A synthetic people stereo patch dataset (S2P2) is introduced to learn wide baseline dense stereo matching for people. The proposed framework not only learns human specific features from synthetic data but also exploits pooling layer and data augmentation to adapt to real data. The network learns from the human specific stereo patches from the proposed dataset for wide-baseline stereo estimation. In addition to patch match learning, a stereo constraint is introduced in the framework to solve wide baseline stereo reconstruction of humans. Quantitative and qualitative performance evaluation against state-of-the-art methods of proposed method demonstrates improved wide baseline stereo reconstruction on challenging datasets. We show that it is possible to learn stereo matching from synthetic people dataset and improve performance on real datasets for stereo reconstruction of people from narrow and wide baseline stereo data.


  • Open Access

    Anomaly based target detection in hyperspectral images via graph cuts

    by Emrecan Bati, Acar Erdinc, Davut Cesmeci, Akin Caliskan, Alper Koz, Selim Aksoy, Sarp Erturk, A. Aydin Alatan

    The studies on hyperspectral target detection until now, has been treated in two approaches. Anomaly detection can be considered as the first approach, which analyses the hyperspectral image with respect to the difference between target and the rest of the hyperspectral image. The second approach compares the previously obtained spectral signature of the target with the pixels of the hyperspectral image in order to localize the target. A distinctive disadvantage of the aforementioned approaches is to treat each pixel of the hyperspectral image individually, without considering the neighbourhood relations between the pixels. In this paper, we propose a target detection algorithm which combines the anomaly detection and signature based hyperspectral target detection approaches in a graph based framework by utilizing the neighbourhood relations between the pixels. Assuming that the target signature is available and the target sizes are in the range of anomaly sizes, a novel derivative based matched filter is first proposed to model the foreground. Second, a new anomaly detection method which models the background as a Gaussian mixture is developed. The developed model estimates the optimal number of components forming the Gaussian mixture by means of utilizing sparsity information. Finally, the similarity of the neighbouring hyperspectral pixels is measured with the spectral angle mapper. The overall proposed graph based method has successfully combined the foreground, background and neighbouring information and improved the detection performance by locating the target as a whole object free from noises.