Boosting Few-Shot Learning With Adaptive Margin Loss

Few-shot learning (FSL) has attracted increasing attention in recent years but remains challenging, due to the intrinsic difficulty in learning to generalize from a few examples. This paper proposes an adaptive margin principle to improve the generalization ability of metric-based meta-learning approaches for few-shot learning problems. Specifically, we first develop a class-relevant additive margin loss, where semantic similarity between each pair of classes is considered to separate samples in the feature embedding space from similar classes. Further, we incorporate the semantic context among all classes in a sampled training task and develop a task-relevant additive margin loss to better distinguish samples from different classes. Our adaptive margin method can be easily extended to a more realistic generalized FSL setting. Extensive experiments demonstrate that the proposed method can boost the performance of current metric-based meta-learning approaches, under both the standard FSL and generalized FSL settings.

Universal Person Re-Identification

Most state-of-the-art person re-identification (re-id) methods depend on supervised model learning with a large set of cross-view identity labelled training data. Even worse, such trained models are limited to only the same-domain deployment with significantly degraded cross-domain generalization capability, i.e. domain specific. To solve this limitation, there are a number of recent unsupervised domain adaptation and unsupervised learning methods that leverage unlabelled target domain training data. However, these methods need to train a separate model for each target domain as supervised learning methods. This conventional {em train once, run once} pattern is unscalable to a large number of target domains typically encountered in real-world deployments. We address this problem by presenting a train once, run everywhere pattern industry-scale systems are desperate for. We formulate a universal model learning approach enabling domain-generic person re-id using only limited training data of a {em single} seed domain. Specifically, we train a universal re-id deep model to discriminate between a set of transformed person identity classes. Each of such classes is formed by applying a variety of random appearance transformations to the images of that class, where the transformations simulate the camera viewing conditions of any domains for making the model training domain generic. Extensive evaluations show the superiority of our method for universal person re-id over a wide variety of state-of-the-art unsupervised domain adaptation and unsupervised learning re-id methods on five standard benchmarks: Market-1501, DukeMTMC, CUHK03, MSMT17, and VIPeR.

Self-Referenced Deep Learning

Knowledge distillation is an effective approach to transferring knowledge from a teacher neural network to a student target network for satisfying the low-memory and fast running requirements in practice use. Whilst being able to create stronger target networks compared to the vanilla non-teacher based learning strategy, this scheme needs to train additionally a large teacher model with expensive computational cost. In this work, we present a Self-Referenced Deep Learning (SRDL) strategy. Unlike both vanilla optimisation and existing knowledge distillation, SRDL distils the knowledge discovered by the in-training target model back to itself to regularise the subsequent learning procedure therefore eliminating the need for training a large teacher model. SRDL improves the model generalisation performance compared to vanilla learning and conventional knowledge distillation approaches with negligible extra computational cost. Extensive evaluations show that a variety of deep networks benefit from SRDL resulting in enhanced deployment performance on both coarse-grained object categorisation tasks (CIFAR10, CIFAR100, Tiny ImageNet, and ImageNet) and fine-grained person instance identification tasks (Market-1501).

Person Search by Multi-Scale Matching

We consider the problem of person search in unconstrained scene images. Existing methods usually focus on improving the person detection accuracy to mitigate negative effects imposed by misalignment, mis-detections, and false alarms resulted from noisy people auto-detection. In contrast to previous studies, we show that sufficiently reliable person instance cropping is achievable by slightly improved state-of-the-art deep learning object detectors (e.g. Faster-RCNN), and the under-studied multi-scale matching problem in person search is a more severe barrier. In this work, we address this multi-scale person search challenge by proposing a Cross-Level Semantic Alignment (CLSA) deep learning approach capable of learning more discriminative identity feature representations in a unified end-to-end model. This is realised by exploiting the in-network feature pyramid structure of a deep neural network enhanced by a novel cross pyramid-level semantic alignment loss function. This favourably eliminates the need for constructing a computationally expensive image pyramid and a complex multi-branch network architecture. Extensive experiments show the modelling advantages and performance superiority of CLSA over the state-of-the-art person search and multi-scale matching methods on two large person search benchmarking datasets: CUHK-SYSU and PRW.

Knowledge Distillation by On-the-Fly Native Ensemble

Knowledge distillation is effective to train small and generalisable network models for meeting the low-memory and fast running requirements. Existing offline distillation methods rely on a strong pre-trained teacher, which enables favourable knowledge discovery and transfer but requires a complex two-phase training procedure. Online counterparts address this limitation at the price of lacking a highcapacity teacher. In this work, we present an On-the-fly Native Ensemble (ONE) strategy for one-stage online distillation. Specifically, ONE trains only a single multi-branch network while simultaneously establishing a strong teacher on-the- fly to enhance the learning of target network. Extensive evaluations show that ONE improves the generalisation performance a variety of deep neural networks more significantly than alternative methods on four image classification dataset: CIFAR10, CIFAR100, SVHN, and ImageNet, whilst having the computational efficiency advantages.

Finite-Size Effects in the Absorption Spectra of a Single-Wall Carbon Nanotube

The determination of the optical spectrum of single-wall carbon nanotubes (SWCNTs) is essential for the development of opto-electronic components and sensors with application in many fields. Real SWCNTs are finite, but almost all the studies performed so far use infinite SWCNTs. However, the spectra of finite and infinite systems are different. In this work the optical spectrum of finite (3,3) and (5,5) SWCNTs is calculated as a function of nanotube length. For the (3,3) SWCNTs, the calculated absorption spectra for light polarised both parallel and perpendicularly to the nanotube axis are in good agreement with experimental results. However, our results indicate that the lowest energy peak present in the experimental results for light polarised parallel to the nanotube axis can be attributed to a surface-plasmon resonance that is a consequence of the finite nature of the SWCNTs and not to the presence of SWCNTs with other chiralities, as claimed by the previous theoretical works. The surface-plasmon resonance is also studied using the Aharonov-Bohm effect. Finally, this work demonstrates that the surface-plasmon resonance in finite SWCNT can be described using a 1D infinite well.

Deep Reinforcement Learning Attention Selection for Person Re-Identification

Existing person re-identification (re-id) methods assume the provision of accurately cropped person bounding boxes with minimum background noise, mostly by manually cropping. This is significantly breached in practice when person bounding boxes must be detected automatically given a very large number of images and/or videos processed. Compared to carefully cropped manually, auto-detected bounding boxes are far less accurate with random amount of background clutter which can degrade notably person re-id matching accuracy. In this work, we develop a joint learning deep model that optimises person re-id attention selection within any auto-detected person bounding boxes by reinforcement learning of background clutter minimisation subject to re-id label pairwise constraints. Specifically, we formulate a novel unified re-id architecture called Identity DiscriminativE Attention reinforcement Learning (IDEAL) to accurately select re-id attention in auto-detected bounding boxes for optimising re-id performance. Our model can improve re-id accuracy comparable to that from exhaustive human manual cropping of bounding boxes with additional advantages from identity discriminative attention selection that specially benefits re-id tasks beyond human knowledge. Extensive comparative evaluations demonstrate the re-id advantages of the proposed IDEAL model over a wide range of state-of-the-art re-id methods on two auto-detected re-id benchmarks CUHK03 and Market-1501.