While the Transformer architecture has become the de-facto standard for natural language processing tasks, its applications to computer vision remain limited. In vision, attention is either applied in conjunction with convolutional networks, or used to replace certain components of convolutional networks while keeping their overall structure in place. We show that this reliance on CNNs is not necessary and a pure transformer applied directly to sequences of image patches can perform very well on image classification tasks. When pre-trained on large amounts of data and transferred to multiple mid-sized or small image recognition benchmarks (ImageNet, CIFAR-100, VTAB, etc.), Vision Transformer (ViT) attains excellent results compared to state-of-the-art convolutional networks while requiring substantially fewer computational resources to train.

An Image is Worth 16x16 Words: Transformers for Image Recognition at Scale

This paper presents SimCLR: a simple framework for contrastive learning of visual representations. We simplify recently proposed contrastive self-supervised learning algorithms without requiring specialized architectures or a memory bank. In order to understand what enables the contrastive prediction tasks to learn useful representations, we systematically study the major components of our framework. We show that (1) composition of data augmentations plays a critical role in defining effective predictive tasks, (2) introducing a learnable nonlinear transformation between the representation and the contrastive loss substantially improves the quality of the learned representations, and (3) contrastive learning benefits from larger batch sizes and more training steps compared to supervised learning. By combining these findings, we are able to considerably outperform previous methods for self-supervised and semi-supervised learning on ImageNet. A linear classifier trained on self-supervised representations learned by SimCLR achieves 76.5% top-1 accuracy, which is a 7% relative improvement over previous state-of-the-art, matching the performance of a supervised ResNet-50. When fine-tuned on only 1% of the labels, we achieve 85.8% top-5 accuracy, outperforming AlexNet with 100X fewer labels.

A Simple Framework for Contrastive Learning of Visual Representations

In modern face recognition, the conventional pipeline consists of four stages: detect => align => represent => classify. We revisit both the alignment step and the representation step by employing explicit 3D face modeling in order to apply a piecewise affine transformation, and derive a face representation from a nine-layer deep neural network. This deep network involves more than 120 million parameters using several locally connected layers without weight sharing, rather than the standard convolutional layers. Thus we trained it on the largest facial dataset to-date, an identity labeled dataset of four million facial images belonging to more than 4, 000 identities. The learned representations coupling the accurate model-based alignment with the large facial database generalize remarkably well to faces in unconstrained environments, even with a simple classifier. Our method reaches an accuracy of 97.35% on the Labeled Faces in the Wild (LFW) dataset, reducing the error of the current state of the art by more than 27%, closely approaching human-level performance.

/pdf/deepface-closing-the-gap-to-human-level-performance-in-face-3wcmadzjn3.pdf

DeepFace: Closing the Gap to Human-Level Performance in Face Verification

The goal of this paper is face recognition – from either a single photograph or from a set of faces tracked in a video. Recent progress in this area has been due to two factors: (i) end to end learning for the task using a convolutional neural network (CNN), and (ii) the availability of very large scale training datasets. We make two contributions: first, we show how a very large scale dataset (2.6M images, over 2.6K people) can be assembled by a combination of automation and human in the loop, and discuss the trade off between data purity and time; second, we traverse through the complexities of deep network training and face recognition to present methods and procedures to achieve comparable state of the art results on the standard LFW and YTF face benchmarks.

/pdf/deep-face-recognition-163wh6e4b2.pdf

Deep face recognition

One of the main challenges in feature learning using Deep Convolutional Neural Networks (DCNNs) for large-scale face recognition is the design of appropriate loss functions that can enhance the discriminative power. Centre loss penalises the distance between deep features and their corresponding class centres in the Euclidean space to achieve intra-class compactness. SphereFace assumes that the linear transformation matrix in the last fully connected layer can be used as a representation of the class centres in the angular space and therefore penalises the angles between deep features and their corresponding weights in a multiplicative way. Recently, a popular line of research is to incorporate margins in well-established loss functions in order to maximise face class separability. In this paper, we propose an Additive Angular Margin Loss (ArcFace) to obtain highly discriminative features for face recognition. The proposed ArcFace has a clear geometric interpretation due to its exact correspondence to geodesic distance on a hypersphere. We present arguably the most extensive experimental evaluation against all recent state-of-the-art face recognition methods on ten face recognition benchmarks which includes a new large-scale image database with trillions of pairs and a large-scale video dataset. We show that ArcFace consistently outperforms the state of the art and can be easily implemented with negligible computational overhead. To facilitate future research, the code has been made available.

/pdf/arcface-additive-angular-margin-loss-for-deep-face-7vvlem659h.pdf

ArcFace: Additive Angular Margin Loss for Deep Face Recognition

The goal of this paper is unsupervised face clustering in edited video material – where face tracks arising from different people are assigned to separate clusters, with one cluster for each person. In particular we explore the extent to which faces can be clustered automatically without making an error. This is a very challenging problem given the variation in pose, lighting and expressions that can occur, and the similarities between different people.The novelty we bring is three fold: first, we show that a form of weak supervision is available from the editing structure of the material – the shots, threads and scenes that are standard in edited video; second, we show that by first clustering within scenes the number of face tracks can be significantly reduced with almost no errors; third, we propose an extension of the clustering method to entire episodes using exemplar SVMs based on the negative training data automatically harvested from the editing structure.The method is demonstrated on multiple episodes from two very different TV series, Scrubs and Buffy. For both series it is shown that we move towards our goal, and also outperform a number of baselines from previous works.

/pdf/total-cluster-a-person-agnostic-clustering-method-for-mbqojx5f44.pdf

Total Cluster: A person agnostic clustering method for broadcast videos

We describe a method of visual search for finding people in large video datasets. The novelty is that the person of interest can be specified at run time by a text query, and a discriminative classifier for that person is then learnt on-the-fly using images downloaded from Google Image search. The performance of the method is evaluated on a ground truth dataset of episodes of Scrubs, and results are also shown for retrieval on the TRECVid 2011 IACC.1.B dataset of over 8k videos. The entire process from specifying the query to receiving the ranked results takes only a matter of seconds.

/pdf/on-the-fly-specific-person-retrieval-38s6ebqqop.pdf

On-the-fly specific person retrieval

The AXES project participated in the interactive instance search task (INS), the known-item search task (KIS), and the multimedia event detection task (MED) for TRECVid 2012. As in our TRECVid 2011 system, we used nearly identical search systems and user interfaces for both INS and KIS. Our interactive INS and KIS systems focused this year on using classifiers trained at query time with positive examples collected from external search engines. Participants in our KIS experiments were media professionals from the BBC; our INS experiments were carried out by students and researchers at Dublin City University. We performed comparatively well in both experiments. Our best KIS run found 13 of the 25 topics, and our best INS runs outperformed all other submitted runs in terms of P@100. For MED, the system presented was based on a minimal number of low-level descriptors, which we chose to be as large as computationally feasible. These descriptors are aggregated to produce high-dimensional video-level signatures, which are used to train a set of linear classifiers. Our MED system achieved the second-best score of all submitted runs in the main track, and best score in the ad-hoc track, suggesting that a simple system based on state-of-the-art low-level descriptors can give relatively high performance. This paper describes in detail our KIS, INS, and MED systems and the results and findings of our experiments.

/pdf/axes-at-trecvid-2012-kis-ins-and-med-2qu83y75br.pdf

AXES at TRECVid 2012: KIS, INS, and MED

The AXES project participated in the interactive instance search task (INS), the semantic indexing task (SIN) the multimedia event recounting task (MER), and the multimedia event detection task (MED) for TRECVid 2013. Our interactive INS focused this year on using classifiers trained at query time with positive examples collected from external search engines. Participants in our INS experiments were carried out by students and researchers at Dublin City University. Our best INS runs performed on par with the top ranked INS runs in terms of P@10 and P@30, and around the median in terms of mAP. For SIN, MED and MER, we use systems based on state-of-the-art local low-level descriptors for motion, image, and sound, as well as high-level features to capture speech and text and the visual and audio stream respectively. The low-level descriptors were aggregated by means of Fisher vectors into high-dimensional video-level signatures, the high-level features are aggregated into bag-of-word histograms. Using these features we train linear classifiers, and use early and late-fusion to combine the different features. Our MED system achieved the best score of all submitted runs in the main track, as well as in the ad-hoc track. This paper describes in detail our INS, MER, and MED systems and the results and findings of our experiments.

/pdf/the-axes-submissions-at-trecvid-2013-4134b815mh.pdf

The AXES submissions at TrecVid 2013

In this paper, we present GrokNet, a deployed image recognition system for commerce applications. GrokNet leverages a multi-task learning approach to train a single computer vision trunk. We achieve a 2.1x improvement in exact product match accuracy when compared to the previous state-of-the-art Facebook product recognition system. We achieve this by training on 7 datasets across several commerce verticals, using 80 categorical loss functions and 3 embedding losses. We share our experience of combining diverse sources with wide-ranging label semantics and image statistics, including learning from human annotations, user-generated tags, and noisy search engine interaction data. GrokNet has demonstrated gains in production applications and operates at Facebook scale.

Omkar M. Parkhi

Papers

Total Cluster: A person agnostic clustering method for broadcast videos

On-the-fly specific person retrieval

AXES at TRECVid 2012: KIS, INS, and MED

The AXES submissions at TrecVid 2013

GrokNet: Unified Computer Vision Model Trunk and Embeddings For Commerce