Search Results for author: Christopher Kanan

Found 59 papers, 23 papers with code

GRASP: A Rehearsal Policy for Efficient Online Continual Learning

no code implementations25 Aug 2023 Md Yousuf Harun, Jhair Gallardo, Christopher Kanan

We evaluate GRASP and other policies by conducting CL experiments on the large-scale ImageNet-1K and Places-LT image classification datasets.

Class Incremental Learning Image Classification +3

Understanding the Benefits of Image Augmentations

no code implementations9 Jun 2023 Matthew Iceland, Christopher Kanan

Image Augmentations are widely used to reduce overfitting in neural networks.

Transfer Learning

Overcoming the Stability Gap in Continual Learning

no code implementations2 Jun 2023 Md Yousuf Harun, Christopher Kanan

Addressing this problem would enable learning new data with fewer network updates, resulting in increased computational efficiency.

Class Incremental Learning Incremental Learning

SIESTA: Efficient Online Continual Learning with Sleep

1 code implementation19 Mar 2023 Md Yousuf Harun, Jhair Gallardo, Tyler L. Hayes, Ronald Kemker, Christopher Kanan

Compared to REMIND and prior arts, SIESTA is far more computationally efficient, enabling continual learning on ImageNet-1K in under 2 hours on a single GPU; moreover, in the augmentation-free setting it matches the performance of the offline learner, a milestone critical to driving adoption of continual learning in real-world applications.

Continual Learning

System Design for an Integrated Lifelong Reinforcement Learning Agent for Real-Time Strategy Games

no code implementations8 Dec 2022 Indranil Sur, Zachary Daniels, Abrar Rahman, Kamil Faber, Gianmarco J. Gallardo, Tyler L. Hayes, Cameron E. Taylor, Mustafa Burak Gurbuz, James Smith, Sahana Joshi, Nathalie Japkowicz, Michael Baron, Zsolt Kira, Christopher Kanan, Roberto Corizzo, Ajay Divakaran, Michael Piacentino, Jesse Hostetler, Aswin Raghavan

In this paper, we introduce the Lifelong Reinforcement Learning Components Framework (L2RLCF), which standardizes L2RL systems and assimilates different continual learning components (each addressing different aspects of the lifelong learning problem) into a unified system.

Continual Learning reinforcement-learning +2

Semantic Segmentation with Active Semi-Supervised Representation Learning

no code implementations16 Oct 2022 Aneesh Rangnekar, Christopher Kanan, Matthew Hoffman

We achieve more than 95% of the network's performance on CamVid and CityScapes datasets, utilizing only 12. 1% and 15. 1% of the labeled data, respectively.

Active Learning Contrastive Learning +4

EllSeg-Gen, towards Domain Generalization for head-mounted eyetracking

1 code implementation4 May 2022 Rakshit S. Kothari, Reynold J. Bailey, Christopher Kanan, Jeff B. Pelz, Gabriel J. Diaz

Previous work has shown that convolutional networks excel at extracting gaze features despite the presence of such artifacts.

Domain Generalization Gaze Estimation

OccamNets: Mitigating Dataset Bias by Favoring Simpler Hypotheses

1 code implementation5 Apr 2022 Robik Shrestha, Kushal Kafle, Christopher Kanan

We propose a new direction: modifying the network architecture to impose inductive biases that make the network robust to dataset bias.

Action Recognition

Online Continual Learning for Embedded Devices

1 code implementation21 Mar 2022 Tyler L. Hayes, Christopher Kanan

Real-time on-device continual learning is needed for new applications such as home robots, user personalization on smartphones, and augmented/virtual reality headsets.

Continual Learning

Can I see an Example? Active Learning the Long Tail of Attributes and Relations

no code implementations11 Mar 2022 Tyler L. Hayes, Maximilian Nickel, Christopher Kanan, Ludovic Denoyer, Arthur Szlam

Using this framing, we introduce an active sampling method that asks for examples from the tail of the data distribution and show that it outperforms classical active learning methods on Visual Genome.

Active Learning

Detecting out-of-context objects using contextual cues

no code implementations11 Feb 2022 Manoj Acharya, Anirban Roy, Kaushik Koneripalli, Susmit Jha, Christopher Kanan, Ajay Divakaran

GCRN consists of two separate graphs to predict object labels based on the contextual cues in the image: 1) a representation graph to learn object features based on the neighboring objects and 2) a context graph to explicitly capture contextual cues from the neighboring objects.

2nd Place Solution for SODA10M Challenge 2021 -- Continual Detection Track

no code implementations25 Oct 2021 Manoj Acharya, Christopher Kanan

In this technical report, we present our approaches for the continual object detection track of the SODA10M challenge.

Autonomous Driving object-detection +3

Disentangling Transfer and Interference in Multi-Domain Learning

no code implementations2 Jul 2021 YiPeng Zhang, Tyler L. Hayes, Christopher Kanan

Humans are incredibly good at transferring knowledge from one domain to another, enabling rapid learning of new tasks.

Transfer Learning

How Does Heterogeneous Label Noise Impact Generalization in Neural Nets?

no code implementations29 Jun 2021 Bidur Khanal, Christopher Kanan

Incorrectly labeled examples, or label noise, is common in real-world computer vision datasets.

An Investigation of Critical Issues in Bias Mitigation Techniques

1 code implementation1 Apr 2021 Robik Shrestha, Kushal Kafle, Christopher Kanan

We introduce a new dataset called Biased MNIST that enables assessment of robustness to multiple bias sources.

Question Answering Test +1

Replay in Deep Learning: Current Approaches and Missing Biological Elements

no code implementations1 Apr 2021 Tyler L. Hayes, Giri P. Krishnan, Maxim Bazhenov, Hava T. Siegelmann, Terrence J. Sejnowski, Christopher Kanan

Replay is the reactivation of one or more neural patterns, which are similar to the activation patterns experienced during past waking experiences.


Self-Supervised Training Enhances Online Continual Learning

no code implementations25 Mar 2021 Jhair Gallardo, Tyler L. Hayes, Christopher Kanan

In continual learning, a system must incrementally learn from a non-stationary data stream without catastrophic forgetting.

Continual Learning Image Classification

Selective Replay Enhances Learning in Online Continual Analogical Reasoning

1 code implementation6 Mar 2021 Tyler L. Hayes, Christopher Kanan

Analogical reasoning tests such as Raven's Progressive Matrices (RPMs) are commonly used to measure non-verbal abstract reasoning in humans, and recently offline neural networks for the RPM problem have been proposed.

Continual Learning Image Classification

Improved Robustness to Open Set Inputs via Tempered Mixup

no code implementations10 Sep 2020 Ryne Roady, Tyler L. Hayes, Christopher Kanan

Supervised classification methods often assume that evaluation data is drawn from the same distribution as training data and that all classes are present for training.

Classification General Classification

RODEO: Replay for Online Object Detection

1 code implementation14 Aug 2020 Manoj Acharya, Tyler L. Hayes, Christopher Kanan

Humans can incrementally learn to do new visual detection tasks, which is a huge challenge for today's computer vision systems.

Class Incremental Learning Incremental Learning +2

Stream-51: Streaming Classification and Novelty Detection from Videos

1 code implementation14 Jun 2020 Ryne Roady, Tyler L. Hayes, Hitesh Vaidya, Christopher Kanan

In this work, we introduce Stream-51, a new dataset for streaming classification consisting of temporally correlated images from 51 distinct object categories and additional evaluation classes outside of the training distribution to test novelty recognition.

Classification General Classification +4

On the Value of Out-of-Distribution Testing: An Example of Goodhart's Law

no code implementations NeurIPS 2020 Damien Teney, Kushal Kafle, Robik Shrestha, Ehsan Abbasnejad, Christopher Kanan, Anton Van Den Hengel

Out-of-distribution (OOD) testing is increasingly popular for evaluating a machine learning system's ability to generalize beyond the biases of a training set.

Model Selection Question Answering +1

Do We Need Fully Connected Output Layers in Convolutional Networks?

no code implementations28 Apr 2020 Zhongchao Qian, Tyler L. Hayes, Kushal Kafle, Christopher Kanan

Traditionally, deep convolutional neural networks consist of a series of convolutional and pooling layers followed by one or more fully connected (FC) layers to perform the final classification.

General Classification

A negative case analysis of visual grounding methods for VQA

1 code implementation ACL 2020 Robik Shrestha, Kushal Kafle, Christopher Kanan

Existing Visual Question Answering (VQA) methods tend to exploit dataset biases and spurious statistical correlations, instead of producing right answers for the right reasons.

Question Answering Visual Grounding +1

AeroRIT: A New Scene for Hyperspectral Image Analysis

2 code implementations17 Dec 2019 Aneesh Rangnekar, Nilay Mokashi, Emmett Ientilucci, Christopher Kanan, Matthew J. Hoffman

We investigate applying convolutional neural network (CNN) architecture to facilitate aerial hyperspectral scene understanding and present a new hyperspectral dataset-AeroRIT-that is large enough for CNN training.

Hyperspectral image analysis Image Super-Resolution +5

Are Out-of-Distribution Detection Methods Effective on Large-Scale Datasets?

no code implementations30 Oct 2019 Ryne Roady, Tyler L. Hayes, Ronald Kemker, Ayesha Gonzales, Christopher Kanan

We found that input perturbation and temperature scaling yield the best performance on large scale datasets regardless of the feature space regularization strategy.

General Classification Image Classification +4

Towards calibrated and scalable uncertainty representations for neural networks

no code implementations28 Oct 2019 Nabeel Seedat, Christopher Kanan

For many applications it is critical to know the uncertainty of a neural network's predictions.

REMIND Your Neural Network to Prevent Catastrophic Forgetting

1 code implementation ECCV 2020 Tyler L. Hayes, Kushal Kafle, Robik Shrestha, Manoj Acharya, Christopher Kanan

While there is neuroscientific evidence that the brain replays compressed memories, existing methods for convolutional networks replay raw images.

Quantization Question Answering +1

RITnet: Real-time Semantic Segmentation of the Eye for Gaze Tracking

2 code implementations1 Oct 2019 Aayush K. Chaudhary, Rakshit Kothari, Manoj Acharya, Shusil Dangi, Nitinraj Nair, Reynold Bailey, Christopher Kanan, Gabriel Diaz, Jeff B. Pelz

Accurate eye segmentation can improve eye-gaze estimation and support interactive computing based on visual attention; however, existing eye segmentation methods suffer from issues such as person-dependent accuracy, lack of robustness, and an inability to be run in real-time.

Gaze Estimation Real-Time Semantic Segmentation +1

Lifelong Machine Learning with Deep Streaming Linear Discriminant Analysis

2 code implementations4 Sep 2019 Tyler L. Hayes, Christopher Kanan

By combining streaming linear discriminant analysis with deep learning, we are able to outperform both incremental batch learning and streaming learning algorithms on both ImageNet ILSVRC-2012 and CORe50, a dataset that involves learning to classify from temporally ordered samples.

BIG-bench Machine Learning Class Incremental Learning

Answering Questions about Data Visualizations using Efficient Bimodal Fusion

1 code implementation5 Aug 2019 Kushal Kafle, Robik Shrestha, Brian Price, Scott Cohen, Christopher Kanan

Chart question answering (CQA) is a newly proposed visual question answering (VQA) task where an algorithm must answer questions about data visualizations, e. g. bar charts, pie charts, and line graphs.

Chart Question Answering Optical Character Recognition +3

Rethinking Continual Learning for Autonomous Agents and Robots

no code implementations2 Jul 2019 German I. Parisi, Christopher Kanan

Continual learning refers to the ability of a biological or artificial system to seamlessly learn from continuous streams of information while preventing catastrophic forgetting, i. e., a condition in which new incoming information strongly interferes with previously learned representations.

Continual Learning Transfer Learning

Gaze-in-wild: A dataset for studying eye and head coordination in everyday activities

no code implementations9 May 2019 Rakshit Kothari, Zhizhuo Yang, Christopher Kanan, Reynold Bailey, Jeff Pelz, Gabriel Diaz

Our approach was to collect a novel, naturalistic, and multimodal dataset of eye+head movements when subjects performed everyday tasks while wearing a mobile eye tracker equipped with an inertial measurement unit and a 3D stereo camera.

Classification General Classification

Challenges and Prospects in Vision and Language Research

no code implementations19 Apr 2019 Kushal Kafle, Robik Shrestha, Christopher Kanan

Language grounded image understanding tasks have often been proposed as a method for evaluating progress in artificial intelligence.

Natural Language Understanding Test

Answer Them All! Toward Universal Visual Question Answering Models

2 code implementations CVPR 2019 Robik Shrestha, Kushal Kafle, Christopher Kanan

Visual Question Answering (VQA) research is split into two camps: the first focuses on VQA datasets that require natural image understanding and the second focuses on synthetic datasets that test reasoning.

Question Answering Test +1

TallyQA: Answering Complex Counting Questions

1 code implementation29 Oct 2018 Manoj Acharya, Kushal Kafle, Christopher Kanan

Most counting questions in visual question answering (VQA) datasets are simple and require no more than object detection.

Object Counting object-detection +3

Memory Efficient Experience Replay for Streaming Learning

1 code implementation16 Sep 2018 Tyler L. Hayes, Nathan D. Cahill, Christopher Kanan

We find that full rehearsal can eliminate catastrophic forgetting in a variety of streaming learning settings, with ExStream performing well using far less memory and computation.

EarthMapper: A Tool Box for the Semantic Segmentation of Remote Sensing Imagery

1 code implementation1 Apr 2018 Ronald Kemker, Utsav B. Gewali, Christopher Kanan

Deep learning continues to push state-of-the-art performance for the semantic segmentation of color (i. e., RGB) imagery; however, the lack of annotated data for many remote sensing sensors (i. e. hyperspectral imagery (HSI)) prevents researchers from taking advantage of this recent success.

Segmentation Segmentation Of Remote Sensing Imagery +2

Low-Shot Learning for the Semantic Segmentation of Remote Sensing Imagery

1 code implementation26 Mar 2018 Ronald Kemker, Ryan Luu, Christopher Kanan

These low-shot learning frameworks will reduce the manual image annotation burden and improve semantic segmentation performance for remote sensing imagery.

Few-Shot Image Classification Few-Shot Learning +6

Continual Lifelong Learning with Neural Networks: A Review

no code implementations21 Feb 2018 German I. Parisi, Ronald Kemker, Jose L. Part, Christopher Kanan, Stefan Wermter

Humans and animals have the ability to continually acquire, fine-tune, and transfer knowledge and skills throughout their lifespan.

Retrieval Transfer Learning

Aerial Spectral Super-Resolution using Conditional Adversarial Networks

no code implementations23 Dec 2017 Aneesh Rangnekar, Nilay Mokashi, Emmett Ientilucci, Christopher Kanan, Matthew Hoffman

In contrast to the spectra of ground based images, aerial spectral images have low spatial resolution and suffer from higher noise interference.

Spectral Super-Resolution Super-Resolution +1

FearNet: Brain-Inspired Model for Incremental Learning

no code implementations ICLR 2018 Ronald Kemker, Christopher Kanan

Arguably, the best method for incremental class learning is iCaRL, but it requires storing training examples for each class, making it challenging to scale.

Audio Classification Incremental Learning

Convolutional Drift Networks for Video Classification

no code implementations3 Nov 2017 Dillon Graham, Seyed Hamed Fatemi Langroudi, Christopher Kanan, Dhireesha Kudithipudi

Analyzing spatio-temporal data like video is a challenging task that requires processing visual and temporal information effectively.

Classification General Classification +2

Measuring Catastrophic Forgetting in Neural Networks

no code implementations7 Aug 2017 Ronald Kemker, Marc McClure, Angelina Abitino, Tyler Hayes, Christopher Kanan

Deep neural networks are used in many state-of-the-art systems for machine perception.

An Analysis of Visual Question Answering Algorithms

no code implementations ICCV 2017 Kushal Kafle, Christopher Kanan

As a result, evaluation scores are inflated and predominantly determined by answering easier questions, making it difficult to compare different methods.

Question Answering Visual Question Answering

Algorithms for Semantic Segmentation of Multispectral Remote Sensing Imagery using Deep Learning

1 code implementation19 Mar 2017 Ronald Kemker, Carl Salvaggio, Christopher Kanan

In this paper, we adapt state-of-the-art DCNN frameworks in computer vision for semantic segmentation for MSI imagery.

object-detection Object Detection +3

High-Resolution Multispectral Dataset for Semantic Segmentation

no code implementations6 Mar 2017 Ronald Kemker, Carl Salvaggio, Christopher Kanan

Unmanned aircraft have decreased the cost required to collect remote sensing imagery, which has enabled researchers to collect high-spatial resolution data from multiple sensor modalities more frequently and easily.

General Classification Semantic Segmentation +1

Robotic Grasp Detection using Deep Convolutional Neural Networks

no code implementations24 Nov 2016 Sulabh Kumra, Christopher Kanan

In this paper, we present a novel robotic grasp detection system that predicts the best grasping pose of a parallel-plate robotic gripper for novel objects using the RGB-D image of the scene.

Robotic Grasping

Answer-Type Prediction for Visual Question Answering

no code implementations CVPR 2016 Kushal Kafle, Christopher Kanan

Recently, algorithms for object recognition and related tasks have become sufficiently proficient that new vision tasks can now be pursued.

Object Recognition Question Answering +3

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