http://repository.iiitd.edu.in/xmlui/handle/123456789/848 Sat, 11 Apr 2026 15:46:14 GMT 2026-04-11T15:46:14Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1636 Channel-graph regularized correlation filter for visual object tracking Tyagi, Arjun; Subramanyam, A V (Advisor) Correlation filter (CF) based tracker often disregard or weakly incorporate the importance of feature channels as well as channel similarity. To address this, we propose a channel-graph regularization correlation filter-based visual object tracker (CGRCF). In our work, we study two-channel regularization methods. First is the channel regularization that determines the vital feature channels. Second is the graph-regularization that increases the probability of assigning similar weights based on the properties of feature channels. The proposed tracker can be efficiently solved in the Fourier domain using ADMM (Alternate Direction Method of Multiplier) and achieves a real-time tracking speed of 28FPS. We conduct extensive experimentation on the TC128, VOT2017 and VOT2019 datasets. The proposed tracker demonstrates promising results and performs better than several state of the art CF trackers as well as end-to-end deep learning trackers. Wed, 01 Jul 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/1636 2020-07-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/1633 Disentangling reconstruction network for unsupervised cross-domain person Re-Identification Jain, Harsh Kumar; Subramanyam, A V (Advisor) Unsupervised cross-domain Person Re-Identi cation (Re-ID) severely su ers from the domain gap. While di erent works address this issue, bridging domain gap with high-level representation is hard as it comprises of entangled information including identity, pose, illumination, and other domain-speci c variations. In this work, we propose a disentangled reconstruction method to ad- dress the domain-shift problem for Re-ID in an unsupervised manner. To this end, we have two major contributions. First, we propose to disentangle identity-related and non-identity related features from person images. We also reconstruct the disentangled features using a decoding layer to increase the generalization capability of identity features. Second, in the target do- main, we explicitly consider the camera style transfer images as a data augmentation to address intra-domain discrepancy and to learn the camera invariant features from the target domain. We demonstrate that the auxiliary tasks of disentanglement and reconstruction are helpful to improve the generalization capability of the model and perform cross Re-ID on unlabeled tar- get domain data. Experimental results on the challenging benchmarks of Market-1501 and DukeMTMC-reID demonstrate that our proposed method achieves competitive performance. Mon, 01 Jun 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/1633 2020-06-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/864 A recommendation system involving human-in-the-loop to improve the quality of ontologies Bhattacharyya, Pramit; Mutharaju, Vijaya Raghava (Advisor) Building an ontology is not only a time-consuming process, but it is also confusing, especially for beginners and the inexperienced. Although ontology developers can take the help of domain experts in building an ontology, they are not readily available in several cases for a variety of reasons. Ontology developers have to grapple with several questions related to the choice of classes, properties, and the axioms that should be included. Apart from this, there are aspects such as modularity and reusability that should be taken care of. From among the thousands of publicly available ontologies and vocabularies such as Linked Open Vocabularies (LOV), it is hard to know the terms (classes and properties) that can be reused in the development of an ontology. A similar problem exists in implementing the right set of ontology design patterns (ODPs) from among the several available. Generally, ontology developers make use of their experience in handling these issues, and the inexperienced ones have a hard time. In order to bridge this gap, we propose a tool named OntoSeer, that monitors the ontology development process and provides suggestions in real-time by interacting with the ontology developers to improve the quality of the ontology under development. It can provide suggestions on the naming conventions to follow, vocabulary to reuse, ODPs to implement, and axioms to be added to the ontology. OntoSeer has been implemented as a Protégé plug-in and is available at github link https://github.com/kracr/ontoseer Wed, 01 Jul 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/864 2020-07-01T00:00:00Z http://repository.iiitd.edu.in/xmlui/handle/123456789/862 Spatio-temporal-history model : a deep learning model to predict traffic speed using public transport data Gola, Nikhil; Goyal, Vikram (Advisor) Traffic speed prediction is one of the challenging task and has many applications. Existing solutions either use crowd-sourced data or sophisticated technologies to perform the task, and hence are costly and unreliable. In this thesis, we propose a machine learning technique that uses the public transport movement data to predict the speed/congestion on a given road segment. Specifically, we use DIMTS buses movement data that comes in the form of GPS trajectories. The technique, we call as STH-Model (Spatio-Temporal-Historical Model), is based on CNN and LSTM models and captures the local spatial dynamics and temporal speed trends for its prediction task. We demonstrate the efficacy of our approach on real-time DIMTS data Mon, 01 Jun 2020 00:00:00 GMT http://repository.iiitd.edu.in/xmlui/handle/123456789/862 2020-06-01T00:00:00Z