Abstract:The performance of a lifelong learning (L3) model degrades when it is trained on a series of tasks, as the geometrical formation of the embedding space changes while learning novel concepts sequentially. The majority of existing L3 approaches operate on a fixed-curvature (e.g., zero-curvature Euclidean) space that is not necessarily suitable for modeling the complex geometric structure of data. Furthermore, the distillation strategies apply constraints directly on low-dimensional embeddings, discouraging the L3 model from learning new concepts by making the model highly stable. To address the problem, we propose a distillation strategy named L3DMC that operates on mixed-curvature spaces to preserve the already-learned knowledge by modeling and maintaining complex geometrical structures. We propose to embed the projected low dimensional embedding of fixed-curvature spaces (Euclidean and hyperbolic) to higher-dimensional Reproducing Kernel Hilbert Space (RKHS) using a positive-definite kernel function to attain rich representation. Afterward, we optimize the L3 model by minimizing the discrepancies between the new sample representation and the subspace constructed using the old representation in RKHS. L3DMC is capable of adapting new knowledge better without forgetting old knowledge as it combines the representation power of multiple fixed-curvature spaces and is performed on higher-dimensional RKHS. Thorough experiments on three benchmarks demonstrate the effectiveness of our proposed distillation strategy for medical image classification in L3 settings. Our code implementation is publicly available at <a class="link-external link-https" href="https://github.com/csiro-robotics/L3DMC" rel="external noopener nofollow">this https URL</a>.

Multi-Level Curriculum for Training A Distortion-Aware Barrel Distortion Rectification Model

A Deep Ordinal Distortion Estimation Approach for Distortion Rectification

Model-Free Distortion Rectification Framework Bridged by Distortion Distribution Map

Model-Aware Pre-Training for Radial Distortion Rectification.

RecRecNet: Rectangling Rectified Wide-Angle Images by Thin-Plate Spline Model and DoF-based Curriculum Learning

QueryCDR: Query-based Controllable Distortion Rectification Network for Fisheye Images

Generative and Discriminative Learning for Distorted Image Restoration.

A Simple Yet Effective Pipeline For Radial Distortion Correction

AIR-CNN: a Lightweight Automatic Image Rectification CNN Used for Barrel Distortion

Training-Free Large Model Priors for Multiple-in-One Image Restoration

Learning Distortion Invariant Representation for Image Restoration from A Causality Perspective

DeepClean: Integrated Distortion Identification and Algorithm Selection for Rectifying Image Corruptions

Model-Free Rectification Via Cascaded Distortion Model and Enhanced Backward Flow Network

DaFIR: Distortion-aware Representation Learning for Fisheye Image Rectification

OIDC-Net: Omnidirectional Image Distortion Correction via Coarse-to-Fine Region Attention

Radial Lens Distortion Correction Using Convolutional Neural Networks Trained with Synthesized Images

Iterative Learning for Distorted Image Restoration

L3DMC: Lifelong Learning using Distillation via Mixed-Curvature Space

Radial Lens Distortion Correction by Adding a Weight Layer with Inverted Foveal Models to Convolutional Neural Networks

Dynamic Contrastive Knowledge Distillation for Efficient Image Restoration

An efficient deep learning-based framework for image distortion correction