Omid Poursaeed
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Omid Poursaeed has a total of 26 co-authors. Here are the top co-authors by number of shared publications:
Serge Belongie
Cornell University
5 shared publicationsJawad A. Salehi
School of Electrical Engineering, Sharif University of Technology, Tehran, Iran
3 shared publicationsVladimir G. Kim
Adobe Research
3 shared publicationsAditya Prakash
Indian Institute of Technology Roorkee
1 shared publicationBharath Hariharan
Cornell University
1 shared publicationBicheng Gao
1 shared publication
Latest Publications
- 7 November 2020
- Published by Springer Nature
- p. 667-683
Coupling Explicit and Implicit Surface Representations for Generative 3D Modeling
- 7 November 2020
- Published by Springer Nature
- p. 667-683
Coupling Explicit and Implicit Surface Representations for Generative 3D Modeling
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1018-1028
Self-Supervised Learning of Point Clouds via Orientation Estimation
Point clouds provide a compact and efficient representation of 3D shapes. While deep neural networks have achieved impressive results on point cloud learning tasks, they require massive amounts of manually labeled data, which can be costly and time-consuming to collect. In this paper, we leverage 3D self-supervision for learning downstream tasks on point clouds with fewer labels. A point cloud can be rotated in infinitely many ways, which provides a rich label-free source for self-supervision. We consider the auxiliary task of predicting rotations that in turn leads to useful features for other tasks such as shape classification and 3D keypoint prediction. Using experiments on ShapeNet and ModelNet, we demonstrate that our approach outperforms the state-of-the-art. Moreover, features learned by our model are complementary to other self-supervised methods and combining them leads to further performance improvement.
- 1 November 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1018-1028
Self-Supervised Learning of Point Clouds via Orientation Estimation
Point clouds provide a compact and efficient representation of 3D shapes. While deep neural networks have achieved impressive results on point cloud learning tasks, they require massive amounts of manually labeled data, which can be costly and time-consuming to collect. In this paper, we leverage 3D self-supervision for learning downstream tasks on point clouds with fewer labels. A point cloud can be rotated in infinitely many ways, which provides a rich label-free source for self-supervision. We consider the auxiliary task of predicting rotations that in turn leads to useful features for other tasks such as shape classification and 3D keypoint prediction. Using experiments on ShapeNet and ModelNet, we demonstrate that our approach outperforms the state-of-the-art. Moreover, features learned by our model are complementary to other self-supervised methods and combining them leads to further performance improvement.
- 1 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 3335-3345
Neural Puppet: Generative Layered Cartoon Characters
We propose a learning based method for generating new animations of a cartoon character given a few example images. Our method is designed to learn from a traditionally animated sequence, where each frame is drawn by an artist, and thus the input images lack any common structure, correspondences, or labels. We express pose changes as a deformation of a layered 2.5D template mesh, and devise a novel architecture that learns to predict mesh deformations matching the template to a target image. This enables us to extract a common low-dimensional structure from a diverse set of character poses. We combine recent advances in differentiable rendering as well as mesh-aware models to successfully align common template even if only a few character images are available during training. In addition to coarse poses, character appearance also varies due to shading, out-of-plane motions, and artistic effects. We capture these subtle changes by applying an image translation network to refine the mesh rendering, providing an end-to-end model to generate new animations of a character with high visual quality. We demonstrate that our generative model can be used to synthesize in-between frames and to create data-driven deformation. Our template fitting procedure outperforms state-of-the-art generic techniques for detecting image correspondences.
- 1 March 2020
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 3335-3345
Neural Puppet: Generative Layered Cartoon Characters
We propose a learning based method for generating new animations of a cartoon character given a few example images. Our method is designed to learn from a traditionally animated sequence, where each frame is drawn by an artist, and thus the input images lack any common structure, correspondences, or labels. We express pose changes as a deformation of a layered 2.5D template mesh, and devise a novel architecture that learns to predict mesh deformations matching the template to a target image. This enables us to extract a common low-dimensional structure from a diverse set of character poses. We combine recent advances in differentiable rendering as well as mesh-aware models to successfully align common template even if only a few character images are available during training. In addition to coarse poses, character appearance also varies due to shading, out-of-plane motions, and artistic effects. We capture these subtle changes by applying an image translation network to refine the mesh rendering, providing an end-to-end model to generate new animations of a character with high visual quality. We demonstrate that our generative model can be used to synthesize in-between frames and to create data-driven deformation. Our template fitting procedure outperforms state-of-the-art generic techniques for detecting image correspondences.
- 23 January 2019
- Published by Springer Nature
- p. 485-497
Deep Fundamental Matrix Estimation Without Correspondences
- 23 January 2019
- Published by Springer Nature
- p. 485-497
Deep Fundamental Matrix Estimation Without Correspondences
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 4422-4431
Generative Adversarial Perturbations
In this paper, we propose novel generative models for creating adversarial examples, slightly perturbed images resembling natural images but maliciously crafted to fool pre-trained models. We present trainable deep neural networks for transforming images to adversarial perturbations. Our proposed models can produce image-agnostic and image-dependent perturbations for targeted and non-targeted attacks. We also demonstrate that similar architectures can achieve impressive results in fooling both classification and semantic segmentation models, obviating the need for hand-crafting attack methods for each task. Using extensive experiments on challenging high-resolution datasets such as ImageNet and Cityscapes, we show that our perturbations achieve high fooling rates with small perturbation norms. Moreover, our attacks are considerably faster than current iterative methods at inference time.
- 1 June 2018
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 4422-4431
Generative Adversarial Perturbations
In this paper, we propose novel generative models for creating adversarial examples, slightly perturbed images resembling natural images but maliciously crafted to fool pre-trained models. We present trainable deep neural networks for transforming images to adversarial perturbations. Our proposed models can produce image-agnostic and image-dependent perturbations for targeted and non-targeted attacks. We also demonstrate that similar architectures can achieve impressive results in fooling both classification and semantic segmentation models, obviating the need for hand-crafting attack methods for each task. Using extensive experiments on challenging high-resolution datasets such as ImageNet and Cityscapes, we show that our perturbations achieve high fooling rates with small perturbation norms. Moreover, our attacks are considerably faster than current iterative methods at inference time.
- 3 April 2018
- Published by Springer Nature in Machine Vision and Applications
- Vol. 29 (4) , 667-676
Vision-based real estate price estimation
- 3 April 2018
- Published by Springer Nature in Machine Vision and Applications
- Vol. 29 (4) , 667-676
Vision-based real estate price estimation
- Top Cited Papers
- 1 July 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636919,p. 1866-1875
Stacked Generative Adversarial Networks
In this paper, we propose a novel generative model named Stacked Generative Adversarial Networks (SGAN), which is trained to invert the hierarchical representations of a bottom-up discriminative network. Our model consists of a top-down stack of GANs, each learned to generate lower-level representations conditioned on higher-level representations. A representation discriminator is introduced at each feature hierarchy to encourage the representation manifold of the generator to align with that of the bottom-up discriminative network, leveraging the powerful discriminative representations to guide the generative model. In addition, we introduce a conditional loss that encourages the use of conditional information from the layer above, and a novel entropy loss that maximizes a variational lower bound on the conditional entropy of generator outputs. We first train each stack independently, and then train the whole model end-to-end. Unlike the original GAN that uses a single noise vector to represent all the variations, our SGAN decomposes variations into multiple levels and gradually resolves uncertainties in the top-down generative process. Based on visual inspection, Inception scores and visual Turing test, we demonstrate that SGAN is able to generate images of much higher quality than GANs without stacking.
- 1 July 2017
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- No. 10636919,p. 1866-1875
Top Cited PapersStacked Generative Adversarial Networks
In this paper, we propose a novel generative model named Stacked Generative Adversarial Networks (SGAN), which is trained to invert the hierarchical representations of a bottom-up discriminative network. Our model consists of a top-down stack of GANs, each learned to generate lower-level representations conditioned on higher-level representations. A representation discriminator is introduced at each feature hierarchy to encourage the representation manifold of the generator to align with that of the bottom-up discriminative network, leveraging the powerful discriminative representations to guide the generative model. In addition, we introduce a conditional loss that encourages the use of conditional information from the layer above, and a novel entropy loss that maximizes a variational lower bound on the conditional entropy of generator outputs. We first train each stack independently, and then train the whole model end-to-end. Unlike the original GAN that uses a single noise vector to represent all the variations, our SGAN decomposes variations into multiple levels and gradually resolves uncertainties in the top-down generative process. Based on visual inspection, Inception scores and visual Turing test, we demonstrate that SGAN is able to generate images of much higher quality than GANs without stacking.
- 12 December 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Analytical studies of fragmented-spectrum multi-level OFDM-CDMA technique in cognitive radio networks
In this paper, we present a multi-user resource allocation framework using fragmented-spectrum synchronous OFDM-CDMA modulation over a frequency-selective fading channel. In particular, given pre-existing communications in the spectrum where the system is operating, a channel sensing and estimation method is used to obtain information of subcarrier availability. Given this information, some real-valued multi-level orthogonal codes, which are orthogonal codes with values of {±1, ±2, ±3, ±4,…}, are provided for emerging new users, i.e., cognitive radio users. Additionally, we have obtained a closed form expression for bit error rate of cognitive radio receivers in terms of detection probability of primary users, CR users' sensing time and CR users' signal to noise ratio. Moreover, simulation results obtained in this paper indicate the precision with which the analytical results have been obtained in modeling the aforementioned system.
- 12 December 2016
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-6
Analytical studies of fragmented-spectrum multi-level OFDM-CDMA technique in cognitive radio networks
In this paper, we present a multi-user resource allocation framework using fragmented-spectrum synchronous OFDM-CDMA modulation over a frequency-selective fading channel. In particular, given pre-existing communications in the spectrum where the system is operating, a channel sensing and estimation method is used to obtain information of subcarrier availability. Given this information, some real-valued multi-level orthogonal codes, which are orthogonal codes with values of {±1, ±2, ±3, ±4,…}, are provided for emerging new users, i.e., cognitive radio users. Additionally, we have obtained a closed form expression for bit error rate of cognitive radio receivers in terms of detection probability of primary users, CR users' sensing time and CR users' signal to noise ratio. Moreover, simulation results obtained in this paper indicate the precision with which the analytical results have been obtained in modeling the aforementioned system.
- 1 March 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 159-163
Cognitive-engined spectrum-fragmented synchronous MC-CDMA based on generalized hadamard codes
In Cognitive Radio (CR), in order to avoid interference in the Primary Users (PU), some subcarriers need to be deactivated. In systems based on Multicarrier Code Division Multiple Access (MC-CDMA), this causes losing orthogonality among different spreading codes, leading to poor Bit Error Rate (BER) performance. The performance of such system can be improved by using Generalized Hadamard Codes (GHC) instead of conventional Hadamard codes which avoids the loss of orthogonality. This is due to the fact that unlike conventional Hadamard codes, GHC are conjectured to exist for any arbitrary length. In this paper, we propose using a novel spreading code, namely, GHC for synchronous MC-CDMA in CR networks. The new spreading codes help the CR systems maintaining the data rate while improving the performance by eliminating the orthogonality loss. Finally, the performance of the MCCDMA system using GHC is evaluated and it is shown that the proposed system outperform the system with conventional Hadamard codes.
- 1 March 2015
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 159-163
Cognitive-engined spectrum-fragmented synchronous MC-CDMA based on generalized hadamard codes
In Cognitive Radio (CR), in order to avoid interference in the Primary Users (PU), some subcarriers need to be deactivated. In systems based on Multicarrier Code Division Multiple Access (MC-CDMA), this causes losing orthogonality among different spreading codes, leading to poor Bit Error Rate (BER) performance. The performance of such system can be improved by using Generalized Hadamard Codes (GHC) instead of conventional Hadamard codes which avoids the loss of orthogonality. This is due to the fact that unlike conventional Hadamard codes, GHC are conjectured to exist for any arbitrary length. In this paper, we propose using a novel spreading code, namely, GHC for synchronous MC-CDMA in CR networks. The new spreading codes help the CR systems maintaining the data rate while improving the performance by eliminating the orthogonality loss. Finally, the performance of the MCCDMA system using GHC is evaluated and it is shown that the proposed system outperform the system with conventional Hadamard codes.
- 1 May 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Resource allocation using fragmented-spectrum synchronous OFDM-CDMA in cognitive radio networks
This paper presents a fragmented-spectrum synchronous OFDM-CDMA modulation and utilize it as secondary users modulation in a cognitive radio-based network to provide high data rate by efficiently exploiting available spectrum bands in a target spectral range while simultaneously offering multiple-access capability. In particular, given preexisting communications in the spectrum where the system is operating, a channel sensing and estimation method is used to obtain information of subcarrier availability. Given this information, some three-level codes are provided for emerging new cognitive radio users. Furthermore, analytical results of the system performance in a typical cognitive radio network are shown.
- 1 May 2014
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1-4
Resource allocation using fragmented-spectrum synchronous OFDM-CDMA in cognitive radio networks
This paper presents a fragmented-spectrum synchronous OFDM-CDMA modulation and utilize it as secondary users modulation in a cognitive radio-based network to provide high data rate by efficiently exploiting available spectrum bands in a target spectral range while simultaneously offering multiple-access capability. In particular, given preexisting communications in the spectrum where the system is operating, a channel sensing and estimation method is used to obtain information of subcarrier availability. Given this information, some three-level codes are provided for emerging new cognitive radio users. Furthermore, analytical results of the system performance in a typical cognitive radio network are shown.