SOSavas Ozkan

Savas Ozkan

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

Savas Ozkan has a total of 23 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Cross-Modal Learning via Adversarial Loss and Covariate Shift for Enhanced Liver Segmentation

    by Savas Ozkan, M. Alper Selver, Bora Baydar, Ali Emre Kavur, Cemre Candemir, Gozde Bozdagi Akar

    Despite the widespread use of deep learning methods for semantic segmentation from single imaging modalities, their performance for exploiting multi-domain data still needs to improve. However, the decision-making process in radiology is often guided by data from multiple sources, such as pre-operative evaluation of living donated liver transplantation donors. In such cases, cross-modality performances of deep models become more important. Unfortunately, the domain-dependency of existing techniques limits their clinical acceptability, primarily confining their performance to individual domains. This issue is further formulated as a multi-source domain adaptation problem, which is an emerging field mainly due to the diverse pattern characteristics exhibited from cross-modality data. This paper presents a novel method that can learn robust representations from unpaired cross-modal (CT-MR) data by encapsulating distinct and shared patterns from multiple modalities. In our solution, the covariate shift property is maintained with structural modifications in our architecture. Also, an adversarial loss is adopted to boost the representation capacity. As a result, sparse and rich representations are obtained. Another superiority of our model is that no information about modalities is needed at the training or inference phase. Tests on unpaired CT and MR liver data obtained from the cross-modality task of the CHAOS grand challenge demonstrate that our approach achieves state-of-the-art results with a large margin in both individual metrics and overall scores.


  • Comparison of semi-automatic and deep learning-based automatic methods for liver segmentation in living liver transplant donors

    by A. Emre Kavur, Naciye Sinem Gezer, Mustafa Baris, Yusuf Sahin, Savas Ozkan, Bora Baydar, Ulas Yuksel, Caglar Kilikcier, Sahin Olut, Gozde Bozdagi Akar, et al.

    To compare the accuracy and repeatability of emerging machine learning based (i.e. deep) automatic segmentation algorithms with those of well-established semi-automatic (interactive) methods for determining liver volume in living liver transplant donors at computerized tomography (CT) imaging. A total of 12 (6 semi-, 6 full-automatic) methods are evaluated. The semi-automatic segmentation algorithms are based on both traditional iterative models including watershed, fast marching, region growing, active contours and modern techniques including robust statistical segmenter and super-pixels. These methods entail some sort of interaction mechanism such as placing initialization seeds on images or determining a parameter range. The automatic methods are based on deep learning and they include three framework templates (DeepMedic, NiftyNet and U-Net) the first two of which are applied with default parameter sets and the last two involve adapted novel model designs. For 20 living donors (6 training and 12 test datasets), a group of imaging scientists and radiologists created ground truths by performing manual segmentations on contrast material-enhanced CT images. Each segmentation is evaluated using five metrics (i.e. volume overlap and relative volume errors, average/RMS/maximum symmetrical surface distances). The results are mapped to a scoring system and a final grade is calculated by taking their average. Accuracy and repeatability were evaluated using slice by slice comparisons and volumetric analysis. Diversity and complementarity are observed through heatmaps. Majority voting and Simultaneous Truth and Performance Level Estimation (STAPLE) algorithms are utilized to obtain the fusion of the individual results. The top four methods are determined to be automatic deep models having 79.63, 79.46 and 77.15 and 74.50 scores. Intra-user score is determined as 95.14. Overall, deep automatic segmentation outperformed interactive techniques on all metrics. The mean volume of liver of ground truth is found to be 1409.93 mL ± 271.28 mL, while it is calculated as 1342.21 mL ± 231.24 mL using automatic and 1201.26 mL ± 258.13 mL using interactive methods, showing higher accuracy and less variation on behalf of automatic methods. The qualitative analysis of segmentation results showed significant diversity and complementarity enabling the idea of using ensembles to obtain superior results. The fusion of automatic methods reached 83.87 with majority voting and 86.20 using STAPLE that are only slightly less than fusion of all methods that achieved 86.70 (majority voting) and 88.74 (STAPLE). Use of the new deep learning based automatic segmentation algorithms substantially increases the accuracy and repeatability for segmentation and volumetric measurements of liver. Fusion of automatic methods based on ensemble approaches exhibits best results almost without any additional time cost due to potential parallel execution of multiple models.


  • Open Access

    Selective word encoding for effective text representation

    by Savaş Özkan, Akin Özkan

    Determining the category of a text document from its semantic content is highly motivated in the literature and it has been extensively studied in various applications. Also, the compact representation of the text is a funda- mental step in achieving precise results for the applications and the studies are generously concentrated to improve its performance. In particular, the studies which exploit the aggregation of word-level representations are the mainstream techniques used in the problem. In this paper, we tackle text representation to achieve high performance in different text classification tasks. Throughout the paper, three critical contributions are presented. First, to encode the word- level representations for each text, we adapt a trainable orderless aggregation algorithm to obtain a more discriminative abstract representation by transforming word vectors to the text-level representation. Second, we propose an effective term-weighting scheme to compute the relative importance of words from the context based on their conjunction with the problem in an end-to-end learning manner. Third, we present a weighted loss function to mitigate the class-imbalance problem between the categories. To evaluate the performance, we collect two distinct datasets as Turkish parliament records (i.e. written speeches of four major political parties including 30731/7683 train and test documents) and newspa- per articles (i.e. daily articles of the columnists including 16000/3200 train and test documents) whose data is available on the web. From the results, the proposed method introduces significant performance improvements to the baseline techniques (i.e. VLAD and Fisher Vector) and achieves 0.823% and 0.878% true prediction accuracies for the party membership and the estimation of the category of articles respectively. The performance validates that the proposed con- tributions (i.e. trainable word-encoding model, trainable term-weighting scheme and weighted loss function) significantly outperform the baselines.

  • Top Cited Papers

    EndNet: Sparse AutoEncoder Network for Endmember Extraction and Hyperspectral Unmixing

    by Savas Ozkan, Berk Kaya, Gozde Bozdagi Akar

    Data acquired from multichannel sensors are a highly valuable asset to interpret the environment for a variety of remote sensing applications. However, low spatial resolution is a critical limitation for previous sensors, and the constituent materials of a scene can be mixed in different fractions due to their spatial interactions. Spectral unmixing is a technique that allows us to obtain the material spectral signatures and their fractions from hyperspectral data. In this paper, we propose a novel endmember extraction and hyperspectral unmixing scheme, so-called EndNet, that is based on a two-staged autoencoder network. This well-known structure is completely enhanced and restructured by introducing additional layers and a projection metric [i.e., spectral angle distance (SAD) instead of inner product] to achieve an optimum solution. Moreover, we present a novel loss function that is composed of a Kullback-Leibler divergence term with SAD similarity and additional penalty terms to improve the sparsity of the estimates. These modifications enable us to set the common properties of endmembers, such as nonlinearity and sparsity for autoencoder networks. Finally, due to the stochastic-gradient-based approach, the method is scalable for large-scale data and it can be accelerated on graphical processing units. To demonstrate the superiority of our proposed method, we conduct extensive experiments on several well-known data sets. The results confirm that the proposed method considerably improves the performance compared to the state-of-the-art techniques in the literature.


  • Open Access

    EndNet: Sparse AutoEncoder Network for Endmember Extraction and Hyperspectral Unmixing

    by Savas Ozkan, Berk Kaya, Gozde Bozdagi Akar

    Data acquired from multi-channel sensors is a highly valuable asset to interpret the environment for a variety of remote sensing applications. However, low spatial resolution is a critical limitation for previous sensors and the constituent materials of a scene can be mixed in different fractions due to their spatial interactions. Spectral unmixing is a technique that allows us to obtain the material spectral signatures and their fractions from hyperspectral data. In this paper, we propose a novel endmember extraction and hyperspectral unmixing scheme, so called \textit{EndNet}, that is based on a two-staged autoencoder network. This well-known structure is completely enhanced and restructured by introducing additional layers and a projection metric (i.e., spectral angle distance (SAD) instead of inner product) to achieve an optimum solution. Moreover, we present a novel loss function that is composed of a Kullback-Leibler divergence term with SAD similarity and additional penalty terms to improve the sparsity of the estimates. These modifications enable us to set the common properties of endmembers such as non-linearity and sparsity for autoencoder networks. Lastly, due to the stochastic-gradient based approach, the method is scalable for large-scale data and it can be accelerated on Graphical Processing Units (GPUs). To demonstrate the superiority of our proposed method, we conduct extensive experiments on several well-known datasets. The results confirm that the proposed method considerably improves the performance compared to the state-of-the-art techniques in literature.


  • A comparison of inpainting techniques in image reanimation

    by Ece Selin Böncü, Savaş Özkan, Gözde Bozdagı Akar

    Inpainting applications include object removal on images and videos, crack filling, error concealment, texture synthesis, where in this paper, its usage for image coherence and perspective emphasis on video frames in 2D image-to-video conversion system is analysed. Besides, the performance of different techniques in object removal and image reconstruction is compared using visual experiments and quality metrics.

  • Open Access

    Cloud Detection From RGB Color Remote Sensing Images With Deep Pyramid Networks

    by Savas Ozkan, Mehmet Efendioglu, Caner Demirpolat

    Cloud detection from remotely observed data is a critical pre-processing step for various remote sensing applications. In particular, this problem becomes even harder for RGB color images, since there is no distinct spectral pattern for clouds, which is directly separable from the Earth surface. In this paper, we adapt a deep pyramid network (DPN) to tackle this problem. For this purpose, the network is enhanced with a pre-trained parameter model at the encoder layer. Moreover, the method is able to obtain accurate pixel-level segmentation and classification results from a set of noisy labeled RGB color images. In order to demonstrate the superiority of the method, we collect and label data with the corresponding cloud/non-cloudy masks acquired from low-orbit Gokturk-2 and RASAT satellites. The experimental results validates that the proposed method outperforms several baselines even for hard cases (e.g. snowy mountains) that are perceptually difficult to distinguish by human eyes.

  • Relaxed Spatio-Temporal Deep Feature Aggregation for Real-Fake Expression Prediction

    by Savas Ozkan, Gozde Bozdagi Akar

    Frame-level visual features are generally aggregated in time with the techniques such as LSTM, Fisher Vectors, NetVLAD etc. to produce a robust video-level representation. We here introduce a learnable aggregation technique whose primary objective is to retain short-time temporal structure between frame-level features and their spatial interdependencies in the representation. Also, it can be easily adapted to the cases where there have very scarce training samples. We evaluate the method on a real-fake expression prediction dataset to demonstrate its superiority. Our method obtains 65% score on the test dataset in the official MAP evaluation and there is only one misclassified decision with the best reported result in the Chalearn Challenge (i.e. 66.7%). Lastly, we believe that this method can be extended to different problems such as action/event recognition in future.


  • Open Access

    Relaxed Spatio-Temporal Deep Feature Aggregation for Real-Fake Expression Prediction

    by Savas Ozkan, Gozde Bozdagi Akar

    Frame-level visual features are generally aggregated in time with the techniques such as LSTM, Fisher Vectors, NetVLAD etc. to produce a robust video-level representation. We here introduce a learnable aggregation technique whose primary objective is to retain short-time temporal structure between frame-level features and their spatial interdependencies in the representation. Also, it can be easily adapted to the cases where there have very scarce training samples. We evaluate the method on a real-fake expression prediction dataset to demonstrate its superiority. Our method obtains 65% score on the test dataset in the official MAP evaluation and there is only one misclassified decision with the best reported result in the Chalearn Challenge (i.e. 66:7%) . Lastly, we believe that this method can be extended to different problems such as action/event recognition in future.

  • Open Access

    Large-Scale Video Search with Efficient Temporal Voting Structure

    by Ersin Esen, Savas Ozkan, Ilkay Atil

    In this work, we propose a fast content-based video querying system for large-scale video search. The proposed system is distinguished from similar works with two major contributions. First contribution is superiority of joint usage of repeated content representation and efficient hashing mechanisms. Repeated content representation is utilized with a simple yet robust feature, which is based on edge energy of frames. Each of the representation is converted into hash code with Hamming Embedding method for further queries. Second contribution is novel queue-based voting scheme that leads to modest memory requirements with gradual memory allocation capability, contrary to complete brute-force temporal voting schemes. This aspect enables us to make queries on large video databases conveniently, even on commodity computers with limited memory capacity. Our results show that the system can respond to video queries on a large video database with fast query times, high recall rate and very low memory and disk requirements.