Search Results for author: Hakan Bilen

Found 47 papers, 23 papers with code

Synthesizing Informative Training Samples with GAN

1 code implementation15 Apr 2022 Bo Zhao, Hakan Bilen

However, traditional GANs generated images are not as informative as the real training samples when being used to train deep neural networks.

Universal Representations: A Unified Look at Multiple Task and Domain Learning

2 code implementations6 Apr 2022 Wei-Hong Li, Xialei Liu, Hakan Bilen

We propose a unified look at jointly learning multiple vision tasks and visual domains through universal representations, a single deep neural network.

cross-domain few-shot learning Image Classification

3D Equivariant Graph Implicit Functions

no code implementations31 Mar 2022 Yunlu Chen, Basura Fernando, Hakan Bilen, Matthias Nießner, Efstratios Gavves

In this work, we address two key limitations of such representations, in failing to capture local 3D geometric fine details, and to learn from and generalize to shapes with unseen 3D transformations.

CAFE: Learning to Condense Dataset by Aligning Features

1 code implementation3 Mar 2022 Kai Wang, Bo Zhao, Xiangyu Peng, Zheng Zhu, Shuo Yang, Shuo Wang, Guan Huang, Hakan Bilen, Xinchao Wang, Yang You

Dataset condensation aims at reducing the network training effort through condensing a cumbersome training set into a compact synthetic one.

Learning Multiple Dense Prediction Tasks from Partially Annotated Data

1 code implementation29 Nov 2021 Wei-Hong Li, Xialei Liu, Hakan Bilen

Despite the recent advances in multi-task learning of dense prediction problems, most methods rely on expensive labelled datasets.

Multi-Task Learning

Not All Relations are Equal: Mining Informative Labels for Scene Graph Generation

no code implementations26 Nov 2021 Arushi Goel, Basura Fernando, Frank Keller, Hakan Bilen

Scene graph generation (SGG) aims to capture a wide variety of interactions between pairs of objects, which is essential for full scene understanding.

Graph Generation Informativeness +2

Dataset Condensation with Distribution Matching

1 code implementation8 Oct 2021 Bo Zhao, Hakan Bilen

Computational cost of training state-of-the-art deep models in many learning problems is rapidly increasing due to more sophisticated models and larger datasets.

Continual Learning Neural Architecture Search

Learning to Annotate Part Segmentation with Gradient Matching

no code implementations ICLR 2022 Yu Yang, Xiaotian Cheng, Hakan Bilen, Xiangyang Ji

In particular, we formulate the annotator learning as the following learning-to-learn problem.

Cross-domain Few-shot Learning with Task-specific Adapters

2 code implementations1 Jul 2021 Wei-Hong Li, Xialei Liu, Hakan Bilen

In this paper, we look at the problem of cross-domain few-shot classification that aims to learn a classifier from previously unseen classes and domains with few labeled samples.

cross-domain few-shot learning

Semi-supervised Viewpoint Estimation with Geometry-aware Conditional Generation

no code implementations2 Apr 2021 Octave Mariotti, Hakan Bilen

There is a growing interest in developing computer vision methods that can learn from limited supervision.

Viewpoint Estimation

Learning Foreground-Background Segmentation from Improved Layered GANs

no code implementations1 Apr 2021 Yu Yang, Hakan Bilen, Qiran Zou, Wing Yin Cheung, Xiangyang Ji

Deep learning approaches heavily rely on high-quality human supervision which is nonetheless expensive, time-consuming, and error-prone, especially for image segmentation task.

Semantic Segmentation Unsupervised Object Segmentation

Universal Representation Learning from Multiple Domains for Few-shot Classification

2 code implementations ICCV 2021 Wei-Hong Li, Xialei Liu, Hakan Bilen

In this paper, we look at the problem of few-shot classification that aims to learn a classifier for previously unseen classes and domains from few labeled samples.

Classification Few-Shot Image Classification +2

Dataset Condensation with Differentiable Siamese Augmentation

1 code implementation16 Feb 2021 Bo Zhao, Hakan Bilen

In many machine learning problems, large-scale datasets have become the de-facto standard to train state-of-the-art deep networks at the price of heavy computation load.

Continual Learning Data Augmentation +2

Reducing Implicit Bias in Latent Domain Learning

no code implementations1 Jan 2021 Lucas Deecke, Timothy Hospedales, Hakan Bilen

A fundamental shortcoming of deep neural networks is their specialization to a single task and domain.

Image Classification

ViewNet: Unsupervised Viewpoint Estimation From Conditional Generation

no code implementations ICCV 2021 Octave Mariotti, Oisin Mac Aodha, Hakan Bilen

Understanding the 3D world without supervision is currently a major challenge in computer vision as the annotations required to supervise deep networks for tasks in this domain are expensive to obtain on a large scale.

Image Reconstruction Self-Supervised Learning +1

Deep Anomaly Detection by Residual Adaptation

no code implementations5 Oct 2020 Lucas Deecke, Lukas Ruff, Robert A. Vandermeulen, Hakan Bilen

Deep anomaly detection is a difficult task since, in high dimensions, it is hard to completely characterize a notion of "differentness" when given only examples of normality.

Anomaly Detection Disentanglement

Knowledge Distillation for Multi-task Learning

2 code implementations14 Jul 2020 Wei-Hong Li, Hakan Bilen

We then learn the multi-task model for minimizing task-specific loss and for producing the same feature with task-specific models.

Knowledge Distillation Multi-Task Learning

Dataset Condensation with Gradient Matching

2 code implementations ICLR 2021 Bo Zhao, Konda Reddy Mopuri, Hakan Bilen

As the state-of-the-art machine learning methods in many fields rely on larger datasets, storing datasets and training models on them become significantly more expensive.

Continual Learning Domain Adaptation +1

Continual Representation Learning for Biometric Identification

1 code implementation8 Jun 2020 Bo Zhao, Shixiang Tang, Dapeng Chen, Hakan Bilen, Rui Zhao

With the explosion of digital data in recent years, continuously learning new tasks from a stream of data without forgetting previously acquired knowledge has become increasingly important.

Continual Learning Knowledge Distillation +1

Latent Domain Learning with Dynamic Residual Adapters

no code implementations1 Jun 2020 Lucas Deecke, Timothy Hospedales, Hakan Bilen

While recent techniques in domain adaptation and multi-domain learning enable the learning of more domain-agnostic features, their success relies on the presence of domain labels, typically requiring manual annotation and careful curation of datasets.

Domain Adaptation Image Classification +1

iDLG: Improved Deep Leakage from Gradients

2 code implementations8 Jan 2020 Bo Zhao, Konda Reddy Mopuri, Hakan Bilen

Particularly, our approach can certainly extract the ground-truth labels as opposed to DLG, hence we name it Improved DLG (iDLG).

Federated Learning

Learning to Impute: A General Framework for Semi-supervised Learning

2 code implementations22 Dec 2019 Wei-Hong Li, Chuan-Sheng Foo, Hakan Bilen

Recent semi-supervised learning methods have shown to achieve comparable results to their supervised counterparts while using only a small portion of labels in image classification tasks thanks to their regularization strategies.

Classification Facial Landmark Detection +2

Injecting Prior Knowledge into Image Caption Generation

no code implementations22 Nov 2019 Arushi Goel, Basura Fernando, Thanh-Son Nguyen, Hakan Bilen

Automatically generating natural language descriptions from an image is a challenging problem in artificial intelligence that requires a good understanding of the visual and textual signals and the correlations between them.

Image Captioning

NormGrad: Finding the Pixels that Matter for Training

no code implementations19 Oct 2019 Sylvestre-Alvise Rebuffi, Ruth Fong, Xu Ji, Hakan Bilen, Andrea Vedaldi

In this paper, we are rather interested by the locations of an image that contribute to the model's training.


Image Deconvolution with Deep Image and Kernel Priors

no code implementations18 Oct 2019 Zhunxuan Wang, Zipei Wang, Qiqi Li, Hakan Bilen

Image deconvolution is the process of recovering convolutional degraded images, which is always a hard inverse problem because of its mathematically ill-posed property.

Denoising Image Deconvolution +1

Deep Multiple Instance Learning with Gaussian Weighting

no code implementations25 Sep 2019 Basura Fernando, Hakan Bilen

The instance representation is shared by both instance classification and weighting streams.

Classification Multiple Instance Learning +1

Unsupervised Learning of Landmarks by Descriptor Vector Exchange

1 code implementation ICCV 2019 James Thewlis, Samuel Albanie, Hakan Bilen, Andrea Vedaldi

Equivariance to random image transformations is an effective method to learn landmarks of object categories, such as the eyes and the nose in faces, without manual supervision.

Unsupervised Facial Landmark Detection

Personalised aesthetics with residual adapters

1 code implementation In Iberian Conference on Pattern Recognition and Image Analysis 2019 Carlos Rodríguez - Pardo, Hakan Bilen

The use of computational methods to evaluate aesthetics in photography has gained interest in recent years due to the popularization of convolutional neural networks and the availability of new annotated datasets.

Recommendation Systems

Self-supervised Learning of Interpretable Keypoints from Unlabelled Videos

no code implementations CVPR 2020 Tomas Jakab, Ankush Gupta, Hakan Bilen, Andrea Vedaldi

We propose KeypointGAN, a new method for recognizing the pose of objects from a single image that for learning uses only unlabelled videos and a weak empirical prior on the object poses.

Facial Landmark Detection Image-to-Image Translation +4

Weakly Supervised Gaussian Networks for Action Detection

no code implementations16 Apr 2019 Basura Fernando, Cheston Tan Yin Chet, Hakan Bilen

Detecting temporal extents of human actions in videos is a challenging computer vision problem that requires detailed manual supervision including frame-level labels.

Action Detection Frame +1

Modelling and unsupervised learning of symmetric deformable object categories

no code implementations NeurIPS 2018 James Thewlis, Hakan Bilen, Andrea Vedaldi

We propose a new approach to model and learn, without manual supervision, the symmetries of natural objects, such as faces or flowers, given only images as input.


Mode Normalization

2 code implementations ICLR 2019 Lucas Deecke, Iain Murray, Hakan Bilen

Normalization methods are a central building block in the deep learning toolbox.

Unsupervised learning of object frames by dense equivariant image labelling

no code implementations NeurIPS 2017 James Thewlis, Hakan Bilen, Andrea Vedaldi

One of the key challenges of visual perception is to extract abstract models of 3D objects and object categories from visual measurements, which are affected by complex nuisance factors such as viewpoint, occlusion, motion, and deformations.

Frame Optical Flow Estimation +1

Universal representations:The missing link between faces, text, planktons, and cat breeds

no code implementations25 Jan 2017 Hakan Bilen, Andrea Vedaldi

With the advent of large labelled datasets and high-capacity models, the performance of machine vision systems has been improving rapidly.

Continual Learning

Action Recognition with Dynamic Image Networks

3 code implementations2 Dec 2016 Hakan Bilen, Basura Fernando, Efstratios Gavves, Andrea Vedaldi

This is a powerful idea because it allows to convert any video to an image so that existing CNN models pre-trained for the analysis of still images can be immediately extended to videos.

Action Recognition Optical Flow Estimation

Self-Supervised Video Representation Learning With Odd-One-Out Networks

no code implementations CVPR 2017 Basura Fernando, Hakan Bilen, Efstratios Gavves, Stephen Gould

On action classification, our method obtains 60. 3\% on the UCF101 dataset using only UCF101 data for training which is approximately 10% better than current state-of-the-art self-supervised learning methods.

Action Classification General Classification +4

Integrated perception with recurrent multi-task neural networks

no code implementations NeurIPS 2016 Hakan Bilen, Andrea Vedaldi

Modern discriminative predictors have been shown to match natural intelligences in specific perceptual tasks in image classification, object and part detection, boundary extraction, etc.

Image Classification

Dynamic Image Networks for Action Recognition

1 code implementation CVPR 2016 Hakan Bilen, Basura Fernando, Efstratios Gavves, Andrea Vedaldi, Stephen Gould

We introduce the concept of dynamic image, a novel compact representation of videos useful for video analysis especially when convolutional neural networks (CNNs) are used.

Action Recognition

Weakly Supervised Deep Detection Networks

2 code implementations CVPR 2016 Hakan Bilen, Andrea Vedaldi

Weakly supervised learning of object detection is an important problem in image understanding that still does not have a satisfactory solution.

Classification Data Augmentation +2

Weakly Supervised Object Detection With Convex Clustering

no code implementations CVPR 2015 Hakan Bilen, Marco Pedersoli, Tinne Tuytelaars

However, as learning appearance and localization are two interconnected tasks, the optimization is not convex and the procedure can easily get stuck in a poor local minimum, the algorithm "misses" the object in some images.

Weakly Supervised Object Detection

Object Classification with Adaptable Regions

no code implementations CVPR 2014 Hakan Bilen, Marco Pedersoli, Vinay P. Namboodiri, Tinne Tuytelaars, Luc van Gool

In classification of objects substantial work has gone into improving the low level representation of an image by considering various aspects such as different features, a number of feature pooling and coding techniques and considering different kernels.

Classification General Classification

Cannot find the paper you are looking for? You can Submit a new open access paper.