Diffusion models are proficient at generating high-quality images. They are however effective only when operating at the resolution used during training. Inference at a scaled resolution leads to repetitive patterns and structural distortions. Retraining at higher resolutions quickly becomes prohibitive. Thus, methods enabling pre-existing diffusion models to operate at flexible test-time resolutions are highly desirable. Previous works suffer from frequent artifacts and often introduce large latency overheads. We propose two simple modules that combine to solve these issues. We introduce a Frequency Modulation (FM) module that leverages the Fourier domain to improve the global structure consistency, and an Attention Modulation (AM) module which improves the consistency of local texture patterns, a problem largely ignored in prior works. Our method, coined FAM diffusion, can seamlessly integrate into any latent diffusion model and requires no additional training. Extensive qualitative results highlight the effectiveness of our method in addressing structural and local artifacts, while quantitative results show state-of-the-art performance. Also, our method avoids redundant inference tricks for improved consistency such as patch-based or progressive generation, leading to negligible latency overheads. FAM diffusion project webpage: https://happy-hsy.github.io/projects/Famdiffusion/

There has been immense progress recently in the visual quality of Stable Diffusion-based Super Resolution (SD-SR). However, deploying large diffusion models on computationally restricted devices such as mobile phones remains impractical due to the large model size and high latency. This is compounded for SR as it often operates at high res (e.g. 4K×3K). In this work, we introduce Edge-SD-SR, the first parameter efficient and low latency diffusion model for image super-resolution. Edge-SD-SR consists of ~169M parameters, including UNet, encoder and decoder, and has a complexity of only ∼ 142 GFLOPs. To maintain a high visual quality on such low compute budget, we introduce a number of training strategies: (i) A novel conditioning mechanism on the low-resolution input, coined bidirectional conditioning, which tailors the SD model for the SR task. (ii) Joint training of the UNet and encoder, while decoupling the encodings of the HR and LR images and using a dedicated schedule. (iii) Finetuning the decoder using the UNet’s output to directly tailor the decoder to the latents obtained at inference time. Edge-SD-SR runs efficiently on device, e.g. it can upscale a 128 × 128 patch to 512 × 512 in 38 msec while running on a Samsung S24 DSP, and of a 512 × 512 to 2, 048 × 2, 048 (requiring 25 model evaluations) in just ∼ 1.1 sec. Furthermore, we show that Edge-SD-SR matches or even outperforms state-of-the-art SR approaches on the most established SR benchmarks.

Contrastively-trained Vision-Language Models (VLMs) like CLIP have become the de facto approach for discriminative vision-language representation learning. However, these models have limited language understanding, often exhibiting a "bag of words" behavior. At the same time, Large Vision-Language Models (LVLMs), which combine vision encoders with LLMs, have been shown to be capable of detailed vision-language reasoning, yet their autoregressive nature renders them less suitable for discriminative tasks.In this work, we propose to combine "the best of both worlds": a new training approach for discriminative fine-tuning of LVLMs that results in strong discriminative and compositional capabilities. Essentially, our approach converts a generative LVLM into a discriminative one, unlocking its capability for powerful image-text discrimination combined with enhanced language understanding.Our contributions include (1) A carefully designed training/optimization framework that utilizes image-text pairs of variable length and granularity for training the model with both contrastive and next-token prediction losses. This is accompanied by ablation studies that justify the necessity of our framework’s components. (2) A parameter-efficient adaptation method using a combination of soft prompting and LoRA adapters. (3) Significant improvements over state-of-the-art CLIP-like models of similar size, including standard image-text retrieval benchmarks and notable gains in compositionality.

Contrastively-trained Vision-Language Models (VLMs), such as CLIP, have become the standard approach for learning discriminative vision-language representations. However, these models often exhibit shallow language understanding, manifesting bag-of-words behaviour. These limitations are reinforced by their dual-encoder design, which induces a modality gap. Additionally, the reliance on vast web-collected data corpora for training makes the process computationally expensive and introduces significant privacy concerns. To address these limitations, in this work, we challenge the necessity of vision encoders for retrieval tasks by introducing a \textit{vision-free, single-encoder} retrieval pipeline. Departing from the traditional text-to-image retrieval paradigm, we migrate to a text-to-text paradigm with the assistance of VLLM-generated structured image descriptions. We demonstrate that this paradigm shift has significant advantages, including a substantial reduction of the modality gap, improved compositionality, and better performance on short and long caption queries, all attainable with only two hours of calibration on two GPUs. Additionally, substituting raw images with textual descriptions introduces a more privacy-friendly alternative for retrieval. To further assess generalisation and address some of the shortcomings of prior compositionality benchmarks, we release two benchmarks derived from Flickr30k and COCO, containing diverse compositional queries made of short captions, which we coin subFlickr and subCOCO. Our vision-free retriever matches and often surpasses traditional multimodal models. Importantly, our approach achieves state-of-the-art zero-shot performance on multiple retrieval and compositionality benchmarks, with models as small as 0.3B parameters.

Video-driven neural face reenactment aims to synthesize realistic facial images that successfully preserve the identity and appearance of a source face, while transferring the target head pose and facial expressions. Existing GAN-based methods suffer from either distortions and visual artifacts or poor reconstruction quality, i.e., the background and several important appearance details, such as hair style/color, glasses and accessories, are not faithfully reconstructed. Recent advances in Diffusion Probabilistic Models (DPMs) enable the generation of high-quality realistic images. To this end, in this paper we present DiffusionAct, a novel method that leverages the photo-realistic image generation of diffusion models to perform neural face reenactment. Specifically, we propose to control the semantic space of a Diffusion Autoencoder (DiffAE), in order to edit the facial pose of the input images, defined as the head pose orientation and the facial expressions. Our method allows one-shot, self, and cross-subject reenactment, without requiring subject-specific fine-tuning. We compare against state-of-the-art GAN-, StyleGAN2-, and diffusion-based methods, showing better or on-par reenactment performance. Project page: https://stelabou.github.io/diffusionact/

Recognising emotions in context involves identifying an individual’s apparent emotions while considering contextual cues from the surrounding scene. Previous approaches to this task have typically designed explicit scene-encoding architectures or incorporated external scene-related information, such as captions. However, these methods often utilise limited contextual information or rely on intricate training pipelines to decouple noise from relevant information. In this work, we leverage the capabilities of Vision-and-Large-Language Models (VLLMs) to enhance in-context emotion classification in a more straightforward manner. Our proposed method follows a simple yet effective two-stage approach. First, we prompt VLLMs to generate natural language descriptions of the subject’s apparent emotion in relation to the visual context. Second, the descriptions, along with the visual input, are used to train a transformer-based architecture that fuses text and visual features before the final classification task. This method not only simplifies the training process but also significantly improves performance. Experimental results demonstrate that the textual descriptions effectively guide the model to constrain the noisy visual input, allowing our fused architecture to outperform individual modalities. Our approach achieves state-of-the-art performance across three datasets, BoLD, EMOTIC, and CAER-S, without bells and whistles. Our code will be made publicly available.

Learning with Noisy Labels (LNL) reduces reliance on high-quality labeled data but often overlooks open-set noise, where noisy samples belong to unknown classes, unlike closed-set noise within known categories.This paper advances LNL by reformulating the problem to incorporate open-set noise through a complete noise transition matrix, enabling a theoretical comparison of its impact on classification error rates against closed-set noise. Our analysis reveals that open-set noise induces smaller error increases, with distinct effects from 'hard' (semantically similar to inliers) and 'easy' (dissimilar) variants. We evaluate entropy-based detection, finding it effective only for easy open-set noise, and propose solutions leveraging vision-language models and self-supervised learning to address hard noise challenges. For empirical validation, we introduce CIFAR100-O, ImageNet-O, and a WebVision open-set test set, enabling robust benchmarking of LNL methods under open-set noise conditions. Recognizing classification accuracy's limitations in capturing model robustness, we advocate out-of-distribution (OOD) detection as a complementary metric. Our theoretical and empirical results highlight the unique challenges of open-set noise, offering new tools and evaluation frameworks to enhance LNL robustness in real-world scenarios.

This work aims to compress the vision tokens of an LVLM into a representation that is simultaneously suitable for (a) generative and (b) discriminative tasks, (c) is nearly lossless, and (d) storage-efficient. To this end, we propose C&C, a novel compression method that leverages the LVLM itself for task-agnostic visual token compression. Unlike prior methods that perform token reduction on-the-fly, our approach offloads computation to a dedicated, upfront indexing stage, effectively decoupling compression from generation. This enables learning more powerful representations for generation during inference. At the core of C&C is a ``double-forward pass'' training strategy. During the first forward pass, the LLM (of the LVLM) creates a bottleneck by compressing the dense visual tokens into a few summary tokens. Subsequently, the second forward pass processes the language instruction(s) alongside the summary tokens, used as a direct replacement for the image ones. The training of C&C is guided by two key losses: an autoregressive loss applied after the second pass that provides a direct optimization objective for reconstructing the original information flow, and a contrastive loss applied after the first pass to bolster the representational strength of the summary tokens, particularly for discriminative tasks. Moreover, we propose stage-specific adapters for further enhancing performance. C&C produces highly informative compressed representations. An in-depth ablation study confirms the efficacy of our approach. For generative tasks, we achieve a 2x higher compression rate without compromising capabilities, setting a new state-of-the-art. For discriminative tasks, we establish new state-of-the-art results on image retrieval and compositionality benchmarks.

Abstract Existing knowledge distillation methods mostly focus on distillation of teacher’s prediction and intermediate activation. However, the structured representation, which arguably is one of the most critical ingredients of deep models, is largely overlooked. In this work, we propose a novel semantic representational distillation (SRD) method dedicated for distilling representational knowledge semantically from a pretrained teacher to a target student. The key idea is that we leverage the teacher’s classifier as a semantic critic for evaluating the representations of both teacher and student and distilling the semantic knowledge with high-order structured information over all feature dimensions. This is accomplished by introducing a notion of cross-network logit computed through passing student’s representation into teacher’s classifier. Further, considering the set of seen classes as a basis for the semantic space in a combinatorial perspective, we scale SRD to unseen classes for enabling effective exploitation of largely available, arbitrary unlabeled training data. At the problem level, this establishes an interesting connection between knowledge distillation with open-set semi-supervised learning (SSL). Extensive experiments show that our SRD outperforms significantly previous state-of-the-art knowledge distillation methods on both coarse object classification and fine face recognition tasks, as well as less studied yet practically crucial binary network distillation. Under more realistic open-set SSL settings we introduce, we reveal that knowledge distillation is generally more effective than existing out-of-distribution sample detection, and our proposed SRD is superior over both previous distillation and SSL competitors. The source code is available at https://github.com/jingyang2017/SRD_ossl .

Despite noise and caption quality having been acknowledged as important factors impacting vision-language contrastive pre-training, in this paper, we show that the full potential of improving the training process by addressing such issues is yet to be realized. Specifically, we firstly study and analyze two issues affecting training: incorrect assignment of negative pairs, and low caption quality and diversity. Then, we devise effective solutions for addressing both problems, which essentially require training with multiple true positive pairs. Finally, we propose training with sigmoid loss to address such a requirement. We show very large gains over the current state-of-the-art for both image recognition (~ +6% on average over 11 datasets) and image retrieval (~ + 19% on Flickr30k and ~ + 15% on MSCOCO).

In this paper, we observe that a straight bipartite matching loss can be applied to the output tokens of a vision transformer. This results in a backbone + MLP architecture that can do both tasks without the need of an extra encoder-decoder head and learnable queries. We show that a single MViTv2-S architecture trained with bipartite matching to perform both tasks surpasses the same MViTv2-S when trained with RoI align on pre-computed bounding boxes. With a careful design of token pooling and the proposed training pipeline, our Bipartite-Matching Vision Transformer model, BMViT, achieves +3 mAP on AVA2.2. w.r.t. the two-stage MViTv2-S counterpart. Code is available at https://github.com/IoannaNti/BMViT Action Localization is a challenging problem that combines detection and recognition tasks, which are often addressed separately. State-of-the-art methods rely on off-the-shelf bounding box detections pre-computed at high resolution, and propose transformer models that focus on the classification task alone. Such two-stage solutions are prohibitive for real-time deployment. On the other hand, single-stage methods target both tasks by devoting part of the network (generally the backbone) to sharing the majority of the work-load, compromising performance for speed. These methods build on adding a DETR head with learnable queries that after cross-and self-attention can be sent to corresponding MLPs for detecting a person's bounding box and action. However, DETR-like architectures are challenging to train and can incur in big complexity.

In this work we focus on learning facial representations that can be adapted to train effective face recognition models, particularly in the absence of labels. Firstly, compared with existing labelled face datasets, a vastly larger magnitude of unlabeled faces exists in the real world. We explore the learning strategy of these unlabeled facial images through self-supervised pretraining to transfer gener-alized face recognition performance. Moreover, motivated by one recent finding, that is, the face saliency area is critical for face recognition, in contrast to utilizing random cropped blocks of images for constructing augmentations in pretraining, we utilize patches localized by extracted facial landmarks. This enables our method - namely LAndmark-based Facial Self-supervised learning (LAFS), to learn key representation that is more critical for face recognition. We also incorporate two landmark-specific augmen-tations which introduce more diversity of landmark information to further regularize the learning. With learned landmark-based facial representations, we further adapt the representation for face recognition with regularization mitigating variations in landmark positions. Our method achieves significant improvement over the state-of-the-art on multiple face recognition benchmarks, especially on more challenging few-shot scenarios. The code is available at https://github.com/szlbiubiubiulLAFS_CVPR2024.

Hai Pham, Isma Hadji, Xinnuo Xu, Ziedune Degutyte, Jay Rainey, Evangelos Kazakos, Afsaneh Fazly, Georgios Tzimiropoulos, Brais Martinez. Proceedings of the 18th Conference of the European Chapter of the Association for Computational Linguistics (Volume 1: Long Papers). 2024.

Current Vision-Language (VL) models owe their success to large-scale pre-training on webcollected data, which in turn requires highcapacity architectures and large compute resources for training.We posit that when the downstream tasks are known in advance, which is in practice common, the pretraining process can be aligned to the downstream domain, leading to more efficient and accurate models, while shortening the pretraining step.To this end, we introduce a domain-aligned pretraining strategy that, without additional data collection, improves the accuracy on a domain of interest, herein, that of human activities, while largely preserving the generalist knowledge.At the core of our approach stands a new LLM-based method that, provided with a simple set of concept seeds, produces a concept hierarchy with high coverage of the target domain.The concept hierarchy is used to filter a large-scale webcrawled dataset and, then, enhance the resulting instances with targeted synthetic labels.We study in depth how to train such approaches and their resulting behavior.We further show generalization to video-based data by introducing a fast adaptation approach for transitioning from a static (image) model to a dynamic one (i.e. with temporal modeling).On the domain of interest, our approach significantly outperforms models trained on up to 60× more samples and between 10 -100× shorter training schedules for image retrieval, video retrieval and action recognition.Code will be released.

This paper is on long-term video understanding where the goal is to recognise human actions over long temporal windows (up to minutes long). In prior work, long temporal context is captured by constructing a long-term memory bank consisting of past and future video features which are then integrated into standard (short-term) video recognition backbones through the use of attention mechanisms. Two well-known problems related to this approach are the quadratic complexity of the attention operation and the fact that the whole feature bank must be stored in memory for inference. To address both issues, we propose an alternative to attention-based schemes which is based on a low-rank approximation of the memory obtained using Singular Value Decomposition. Our scheme has two advantages: (a) it reduces complexity by more than an order of magnitude, and (b) it is amenable to an efficient implementation for the calculation of the memory bases in an incremental fashion which does not require the storage of the whole feature bank in memory. The proposed scheme matches or surpasses the accuracy achieved by attention-based mechanisms while being memory-efficient. Through extensive experiments, we demonstrate that our framework generalises to different architectures and tasks, outperforming the state-of-the-art in three datasets.

Learning with Noisy labels (LNL) poses a significant challenge for the Machine Learning community. Some of the most widely used approaches that select as clean samples for which the model itself (the in-training model) has high confidence, e.g., 'small loss', can suffer from the so called 'self-confirmation' bias. This bias arises because the in-training model, is at least partially trained on the noisy labels. Furthermore, in the classification case, an additional challenge arises because some of the label noise is between classes that are visually very similar ('hard noise'). This paper addresses these challenges by proposing a method (CLIPCleaner) that leverages CLIP, a powerful Vision-Language (VL) model for constructing a zero-shot classifier for efficient, offline, clean sample selection. This has the advantage that the sample selection is decoupled from the in-training model and that the sample selection is aware of the semantic and visual similarities between the classes due to the way that CLIP is trained. We provide theoretical justifications and empirical evidence to demonstrate the advantages of CLIP for LNL compared to conventional pre-trained models. Compared to current methods that combine iterative sample selection with various techniques, CLIPCleaner offers a simple, single-step approach that achieves competitive or superior performance on benchmark datasets. To the best of our knowledge, this is the first time a VL model has been used for sample selection to address the problem of Learning with Noisy Labels (LNL), highlighting their potential in the domain.

Current post-training quantization methods for LLMs compress the weights down to 4-bits, with moderate to low degradation in accuracy. However, further reducing the number of bits or accelerating the network while avoiding large accuracy drops, especially for smaller, sub 7B models, remains an actively researched and open problem. To address this, in this work, we introduce Quantization with Binary Bases (QBB), a new approach for low-bit quantization that effectively removes (nearly) all multiplications, reducing the implementation to summations. Our novel approach works by decomposing the original weights into a set of binary (1-bit) matrices using an iterative process. For a given layer, starting from a weight matrix, we first construct an initial approximation using an analytical solution, where each new binary matrix, paired with a scaling vector, approximates the residual error of the previous estimation. Secondly, using gradient descent and a progressive learning curriculum, we find the optimal set of binary matrices and scaling vectors that minimize the $\ell_2$ distance between the produced approximation and original weights. Thirdly, as previous steps are input agnostic, we holistically optimize the scaling vectors alone, calibrating them in student-teacher fashion, with the teacher providing both the data, by autoregressive generation starting from a random token, and the target logits. When evaluated across multiple LLM families, our approach matches and outperforms all prior works, setting a new state-of-the-art result using a summation-only based approach.

Privacy issue is a main concern in developing face recognition techniques. Although synthetic face images can partially mitigate potential legal risks while maintaining effective face recognition (FR) performance, FR models trained by face images synthesized by existing generative approaches frequently suffer from performance degradation problems due to the insufficient discriminative quality of these synthesized samples. In this paper, we systematically investigate what contributes to solid face recognition model training, and reveal that face images with certain degree of similarities to their identity centers show great effectiveness in the performance of trained FR models. Inspired by this, we propose a novel diffusion-based approach (namely **Ce**nter-based Se**mi**-hard Synthetic Face Generation (**CemiFace**) which produces facial samples with various levels of similarity to the subject center, thus allowing to generate face datasets containing effective discriminative samples for training face recognition. Experimental results show that with a modest degree of similarity, training on the generated dataset can produce competitive performance compared to previous generation methods. The code will be available at:https://github.com/szlbiubiubiu/CemiFace

Abstract In this paper, we present our framework for neural face/head reenactment whose goal is to transfer the 3D head orientation and expression of a target face to a source face. Previous methods focus on learning embedding networks for identity and head pose/expression disentanglement which proves to be a rather hard task, degrading the quality of the generated images. We take a different approach, bypassing the training of such networks, by using (fine-tuned) pre-trained GANs which have been shown capable of producing high-quality facial images. Because GANs are characterized by weak controllability, the core of our approach is a method to discover which directions in latent GAN space are responsible for controlling head pose and expression variations. We present a simple pipeline to learn such directions with the aid of a 3D shape model which, by construction, inherently captures disentangled directions for head pose, identity, and expression. Moreover, we show that by embedding real images in the GAN latent space, our method can be successfully used for the reenactment of real-world faces. Our method features several favorable properties including using a single source image (one-shot) and enabling cross-person reenactment. Extensive qualitative and quantitative results show that our approach typically produces reenacted faces of notably higher quality than those produced by state-of-the-art methods for the standard benchmarks of VoxCeleb1 & 2.

Abstract Soft prompt learning has emerged as a promising direction for adapting V &L models to a downstream task using a few training examples. However, current methods significantly overfit the training data suffering from large accuracy degradation when tested on unseen classes from the same domain. In addition, all prior methods operate exclusively under the assumption that both vision and language data is present. To this end, we make the following 5 contributions: (1) To alleviate base class overfitting, we propose a novel Language-Aware Soft Prompting (LASP) learning method by means of a text-to-text cross-entropy loss that maximizes the probability of the learned prompts to be correctly classified with respect to pre-defined hand-crafted textual prompts. (2) To increase the representation capacity of the prompts, we also propose grouped LASP where each group of prompts is optimized with respect to a separate subset of textual prompts. (3) Moreover, we identify a visual-language misalignment introduced by prompt learning and LASP, and more importantly, propose a re-calibration mechanism to address it. (4) Importantly, we show that LASP is inherently amenable to including, during training, virtual classes, i.e. class names for which no visual samples are available, further increasing the robustness of the learned prompts. Expanding for the first time the setting to language-only adaptation, (5) we present a novel zero-shot variant of LASP where no visual samples at all are available for the downstream task. Through evaluations on 11 datasets, we show that our approach (a) significantly outperforms all prior works on soft prompting, and (b) matches and surpasses, for the first time, the accuracy on novel classes obtained by hand-crafted prompts and CLIP for 8 out of 11 test datasets. Finally, (c) we show that our zero-shot variant improves upon CLIP without requiring any extra data. Code will be made available.

Despite the large progress in supervised learning with neural networks, there are significant challenges in obtaining high-quality, large-scale and accurately labelled datasets. In such contexts, how to learn in the presence of noisy labels has received more and more attention. Addressing this relatively intricate problem to attain competitive results predominantly involves designing mechanisms that select samples that are expected to have reliable annotations. However, these methods typically involve multiple off-the-shelf techniques, resulting in intricate structures. Furthermore, they frequently make implicit or explicit assumptions about the noise modes/ratios within the dataset. Such assumptions can compromise model robustness and limit its performance under varying noise conditions. Unlike these methods, in this work, we propose an efficient and effective framework with minimal hyperparameters that achieves SOTA results in various benchmarks. Specifically, we design an efficient and concise training framework consisting of a subset expansion module responsible for exploring non-selected samples and a model training module to further reduce the impact of noise, called NoiseBox. Moreover, diverging from common sample selection methods based on the “small loss” mechanism, we introduce a novel sample selection method based on the neighbouring relationships and label consistency in the feature space. Without bells and whistles, such as model co-training, self-supervised pre-training and semi-supervised learning, and with robustness concerning the settings of its few hyper-parameters, our method significantly surpasses previous methods on both CIFAR10/CIFAR100 with synthetic noise and real-world noisy datasets such as Red Mini-ImageNet, WebVision, Clothing1M and ANIMAL-10N.

Soft prompt learning has recently emerged as one of the methods of choice for adapting V&L models to a downstream task using a few training examples. However, current methods significantly overfit the training data, suffering from large accuracy degradation when tested on unseen classes from the same domain. To this end, in this paper, we make the following 4 contributions: (1) To alleviate base class overfitting, we propose a novel Language- Aware Soft Prompting (LASP) learning method by means of a text-to-text cross-entropy loss that maximizes the probability of the learned prompts to be correctly classified with respect to pre-defined hand-crafted textual prompts. (2) To increase the representation capacity of the prompts, we propose grouped LASP where each group of prompts is optimized with respect to a separate subset of textual prompts. (3) We identify a visual-language misalignment introduced by prompt learning and LASP, and more importantly, propose a re-calibration mechanism to address it. (4) We show that LASP is inherently amenable to including, during training, virtual classes, i.e. class names for which no visual samples are available, further increasing the robustness of the learned prompts. Through evaluations on 11 datasets, we show that our approach (a) significantly outperforms all prior works on soft prompting, and (b) matches and surpasses, for the first time, the accuracy on novel classes obtained by hand-crafted prompts and CLIP for 8 out of 11 test datasets. Code will be made available here.

Clustering has been a major research topic in the field of machine learning, one to which Deep Learning has recently been applied with significant success. However, an aspect of clustering that is not addressed by existing deep clustering methods, is that of efficiently producing multiple, diverse partitionings for a given dataset. This is particularly important, as a diverse set of base clusterings are necessary for consensus clustering, which has been found to produce better and more robust results than relying on a single clustering. To address this gap, we propose Div-Clust, a diversity controlling loss that can be incorporated into existing deep clustering frameworks to produce multiple clusterings with the desired degree of diversity. We conduct experiments with multiple datasets and deep clustering frameworks and show that: a) our method effectively controls diversity across frameworks and datasets with very small additional computational cost, b) the sets of clusterings learned by DivClust include solutions that significantly outperform single-clustering baselines, and c) using an off-the-shelf consensus clustering algorithm, DivClust produces consensus clustering solutions that consistently outperform single-clustering baselines, effectively improving the performance of the base deep clustering framework. Code is available at https://github.com/ManiadisG/DivClust.

This paper is on the problem of Knowledge-Based Visual Question Answering (KB-VQA).Recent works have emphasized the significance of incorporating both explicit (through external databases) and implicit (through LLMs) knowledge to answer questions requiring external knowledge effectively.A common limitation of such approaches is that they consist of relatively complicated pipelines and often heavily rely on accessing GPT-3 API.Our main contribution in this paper is to propose a much simpler and readily reproducible pipeline which, in a nutshell, is based on efficient in-context learning by prompting LLaMA (1 and 2) using question-informative captions as contextual information.Contrary to recent approaches, our method is training-free, does not require access to external databases or APIs, and yet achieves state-of-the-art accuracy on the OK-VQA and A-OK-VQA datasets.Finally, we perform several ablation studies to understand important aspects of our method.

In this paper we address the problem of neural face reenactment, where, given a pair of a source and a target facial image, we need to transfer the target's pose (defined as the head pose and its facial expressions) to the source image, by preserving at the same time the source's identity characteristics (e.g., facial shape, hair style, etc), even in the challenging case where the source and the target faces belong to different identities. In doing so, we address some of the limitations of the state-of-the-art works, namely, a) that they depend on paired training data (i.e., source and target faces have the same identity), b) that they rely on labeled data during inference, and c) that they do not preserve identity in large head pose changes. More specifically, we propose a framework that, using unpaired randomly generated facial images, learns to disentangle the identity characteristics of the face from its pose by incorporating the recently introduced style space S [1] of StyleGAN2 [2], a latent representation space that exhibits remarkable disentanglement properties. By capitalizing on this, we learn to successfully mix a pair of source and target style codes using supervision from a 3D model. The resulting latent code, that is subsequently used for reenactment, consists of latent units corresponding to the facial pose of the target only and of units corresponding to the identity of the source only, leading to notable improvement in the reenactment performance compared to recent state-of-the-art methods. In comparison to state of the art, we quantitatively and qualitatively show that the proposed method produces higher quality results even on extreme pose variations. Finally, we report results on real images by first embedding them on the latent space of the pretrained generator. We make the code and the pretrained models publicly available at: https://github.com/StelaBou/StyleMask.

In this paper, we present our method for neural face reenactment, called HyperReenact, that aims to generate realistic talking head images of a source identity, driven by a target facial pose. Existing state-of-the-art face reenactment methods train controllable generative models that learn to synthesize realistic facial images, yet producing reenacted faces that are prone to significant visual artifacts, especially under the challenging condition of extreme head pose changes, or requiring expensive few-shot fine-tuning to better preserve the source identity characteristics. We propose to address these limitations by leveraging the photorealistic generation ability and the disentangled properties of a pretrained StyleGAN2 generator, by first inverting the real images into its latent space and then using a hypernetwork to perform: (i) refinement of the source identity characteristics and (ii) facial pose re-targeting, eliminating this way the dependence on external editing methods that typically produce artifacts. Our method operates under the one-shot setting (i.e., using a single source frame) and allows for cross-subject reenactment, without requiring any subject-specific fine-tuning. We compare our method both quantitatively and qualitatively against several state-of-the-art techniques on the standard benchmarks of VoxCeleb1 and VoxCeleb2, demonstrating the superiority of our approach in producing artifact-free images, exhibiting remarkable robustness even under extreme head pose changes. We make the code and the pretrained models publicly available at: https://github.com/StelaBou/HyperReenact.

This paper is on Few-Shot Object Detection (FSOD), where given a few templates (examples) depicting a novel class (not seen during training), the goal is to detect all of its occurrences within a set of images. From a practical perspective, an FSOD system must fulfil the following desiderata: (a) it must be used as is, without requiring any fine-tuning at test time, (b) it must be able to process an arbitrary number of novel objects concurrently while supporting an arbitrary number of examples from each class and (c) it must achieve accuracy comparable to a closed system. Towards satisfying (a)-(c), in this work, we make the following contributions: We introduce, for the first time, a simple, yet powerful, few-shot detection transformer (FS-DETR) based on visual prompting that can address both desiderata (a) and (b). Our system builds upon the DETR framework, extending it based on two key ideas: (1) feed the provided visual templates of the novel classes as visual prompts during test time, and (2) "stamp" these prompts with pseudo-class embeddings (akin to soft prompting), which are then predicted at the output of the decoder. Importantly, we show that our system is not only more flexible than existing methods, but also, it makes a step towards satisfying desideratum (c). Specifically, it is significantly more accurate than all methods that do not require fine-tuning and even matches and outperforms the current state-of-the-art fine-tuning based methods on the most well-established benchmarks (PASCAL VOC & MSCOCO).

This paper sets out to solve the following problem: How can we turn a generative video captioning model into an open-world video/action classification model? Video captioning models can naturally produce open-ended free-form descriptions of a given video which, however, might not be discriminative enough for video/action recognition. Unfortunately, when fine-tuned to auto-regress the class names directly, video captioning models overfit the base classes losing their open-world zero-shot capabilities. To alleviate base class overfitting, in this work, we propose to use reinforcement learning to enforce the output of the video captioning model to be more class-level discriminative. Specifically, we propose ReGen, a novel reinforcement learning based framework with a three-fold objective and reward functions: (1) a class-level discrimination reward that enforces the generated caption to be correctly classified into the corresponding action class, (2) a CLIP reward that encourages the generated caption to continue to be descriptive of the input video (i.e. video-specific), and (3) a grammar reward that preserves the grammatical correctness of the caption. We show that ReGen can train a model to produce captions that are: discriminative, video-specific and grammatically correct. Importantly, when evaluated on standard benchmarks for zero- and few-shot action classification, ReGen significantly outperforms the previous state-of-the-art.

Foundational image-language models have generated considerable interest due to their efficient adaptation to downstream tasks by prompt learning. Prompt learning treats part of the language model input as trainable while freezing the rest, and optimizes an Empirical Risk Mini-mization objective. However, Empirical Risk Minimization is known to suffer from distributional shifts which hurt gen-eralizability to prompts unseen during training. By leveraging the regularization ability of Bayesian methods, we frame prompt learning from the Bayesian perspective and formulate it as a variational inference problem. Our approach regularizes the prompt space, reduces overfitting to the seen prompts and improves the prompt generalization on unseen prompts. Our framework is implemented by modeling the input prompt space in a probabilistic manner, as an a priori distribution which makes our proposal compatible with prompt learning approaches that are unconditional or conditional on the image. We demonstrate empirically on 15 benchmarks that Bayesian prompt learning provides an appropriate coverage of the prompt space, prevents learning spurious features, and exploits transferable invariant features. This results in better generalization of unseen prompts, even across different datasets and domains.Code available at: https://github.com/saic-fi/Bayesian-Prompt-Learning

Vision-Language (V-L) models trained with contrastive learning to align the visual and language modalities have been shown to be strong few-shot learners. Soft prompt learning is the method of choice for few-shot downstream adaption aiming to bridge the modality gap caused by the distribution shift induced by the new domain. While parameter-efficient, prompt learning still requires access to the model weights and can be computationally infeasible for large models with billions of parameters. To address these shortcomings, in this work, we describe a black-box method for V-L few-shot adaptation that (a) operates on pre-computed image and text features and hence works without access to the model’s weights, (b) it is orders of magnitude faster at training time, (c) it is amenable to both supervised and unsupervised training, and (d) it can be even used to align image and text features computed from uni-modal models. To achieve this, we propose Linear Feature Alignment (LFA), a simple linear approach for V-L re-alignment in the target domain. LFA is initialized from a closed-form solution to a least-squares problem and then it is iteratively updated by minimizing a re-ranking loss. Despite its simplicity, our approach can even surpass soft-prompt learning methods as shown by extensive experiments on 11 image and 2 video datasets.Code available at: https://github.com/saic-fi/LFA

This paper proposes a novel paradigm for the unsupervised learning of object landmark detectors. Contrary to existing methods that build on auxiliary tasks such as image generation or equivariance, we propose a self-training approach where, departing from generic keypoints, a landmark detector and descriptor is trained to improve itself, tuning the keypoints into distinctive landmarks. To this end, we propose an iterative algorithm that alternates between producing new pseudo-labels through feature clustering and learning distinctive features for each pseudo-class through contrastive learning. With a shared backbone for the landmark detector and descriptor, the keypoint locations progressively converge to stable landmarks, filtering those less stable. Compared to previous works, our approach can learn points that are more flexible in terms of capturing large viewpoint changes. We validate our method on a variety of difficult datasets, including LS3D, BBCPose, Human3.6M and PennAction, achieving new state of the art results. Code and models can be found at https://github.com/dimitrismallis/KeypointsToLandmarks/.

Action Units (AUs) are geometrically-based atomic facial muscle movements known to produce appearance changes at specific facial locations. Motivated by this observation we propose a novel AU modelling problem that consists of jointly estimating their localisation and intensity. To this end, we propose a simple yet efficient approach based on Heatmap Regression that merges both problems into a single task. A Heatmap models whether an AU occurs or not at a given spatial location. To accommodate the joint modelling of AUs intensity, we propose variable size heatmaps, with their amplitude and size varying according to the labelled intensity. Using Heatmap Regression, we can inherit from the progress recently witnessed in facial landmark localisation. Building upon the similarities between both problems, we devise a transfer learning approach where we exploit the knowledge of a network trained on large-scale facial landmark datasets. In particular, we explore different alternatives for transfer learning through a) fine-tuning, b) adaptation layers, c) attention maps, and d) reparametrisation. Our approach effectively inherits the rich facial features produced by a strong face alignment network, with minimal extra computational cost. We empirically validate that our system sets a new state-of-the-art on three popular datasets, namely BP4D, DISFA, and FERA2017.

This article aims to solve two important issues that frequently occur in existing automatic personality analysis systems: 1. Attempting to use very short video segments or even single frames, rather than long-term behaviour, to infer personality traits; 2. Lack of methods to encode person-specific facial dynamics for personality recognition. To deal with these issues, this paper first proposes a novel Rank Loss which utilizes the natural temporal evolution of facial actions, rather than personality labels, for self-supervised learning of facial dynamics. Our approach first trains a generic U-net style model that can infer general facial dynamics learned from a set of unlabelled face videos. Then, the generic model is frozen, and a set of intermediate filters are incorporated into this architecture. The self-supervised learning is then resumed with only person-specific videos. This way, the learned filters’ weights are person-specific, making them a valuable source for modeling person-specific facial dynamics. We then propose to concatenate the weights of the learned filters as a person-specific representation, which can be directly used to predict the personality traits without needing other parts of the network. We evaluate the proposed approach on both self-reported personality and apparent personality datasets. In addition to achieving promising results in the estimation of personality trait scores from videos, we show that the tasks conducted by the subject in the video matters, that fusion of a combination of tasks reaches highest accuracy, and that multi-scale dynamics are more informative than single-scale dynamics.

Temporal context is key to the recognition of expressions of emotion. Existing methods, that rely on recurrent or self-attention models to enforce temporal consistency, work on the feature level, ignoring the task-specific temporal dependencies, and fail to model context uncertainty. To alleviate these issues, we build upon the framework of Neural Processes to propose a method for apparent emotion recognition with three key novel components: (a) probabilistic contextual representation with a global latent variable model; (b) temporal context modelling using task-specific predictions in addition to features; and (c) smart temporal context selection. We validate our approach on four databases, two for Valence and Arousal estimation (SEWA and AffWild2), and two for Action Unit intensity estimation (DISFA and BP4D). Results show a consistent improvement over a series of strong baselines as well as over state-of-the-art methods.

An important component of unsupervised learning by instance-based discrimination is a memory bank for storing a feature representation for each training sample in the dataset. In this paper, we introduce 3 improvements to the vanilla memory bank-based formulation which brings massive accuracy gains: (a) Large mini-batch: we pull multiple augmentations for each sample within the same batch and show that this leads to better models and enhanced memory bank updates. (b) Consistency: we enforce the logits obtained by different augmentations of the same sample to be close without trying to enforce discrimination with respect to negative samples as proposed by previous approaches. (c) Hard negative mining: since instance discrimination is not meaningful for samples that are too visually similar, we devise a novel nearest neighbour approach for improving the memory bank that gradually merges extremely similar data samples that were previously forced to be apart by the instance level classification loss. Overall, our approach greatly improves the vanilla memory-bank based instance discrimination and outperforms all existing methods for both seen and unseen testing categories with cosine similarity.

Mixed-precision networks allow for a variable bit-width quantization for every layer in the network. A major limitation of existing work is that the bit-width for each layer must be predefined during training time. This allows little flexibility if the characteristics of the device on which the network is deployed change during runtime. In this work, we propose Bit-Mixer, the very first method to train a meta-quantized network where during test time any layer can change its bit-width without affecting at all the overall network’s ability for highly accurate inference. To this end, we make 2 key contributions: (a) Transitional Batch-Norms, and (b) a 3-stage optimization process which is shown capable of training such a network. We show that our method can result in mixed precision networks that exhibit the desirable flexibility properties for on-device deployment without compromising accuracy. Code will be made available.

This work addresses the problem of discovering, in an unsupervised manner, interpretable paths in the latent space of pretrained GANs, so as to provide an intuitive and easy way of controlling the underlying generative factors. In doing so, it addresses some of the limitations of the state-of-the-art works, namely, a) that they discover directions that are independent of the latent code, i.e., paths that are linear, and b) that their evaluation relies either on visual inspection or on laborious human labeling. More specifically, we propose to learn non-linear warpings on the latent space, each one parametrized by a set of RBF-based latent space warping functions, and where each warping gives rise to a family of non-linear paths via the gradient of the function. Building on the work of [34], that discovers linear paths, we optimize the trainable parameters of the set of RBFs, so as that images that are generated by codes along different paths, are easily distinguishable by a discriminator network. This leads to easily distinguishable image transformations, such as pose and facial expressions in facial images. We show that linear paths can be derived as a special case of our method, and show experimentally that non-linear paths in the latent space lead to steeper, more disentangled and interpretable changes in the image space than in state-of-the art methods, both qualitatively and quantitatively. We make the code and the pretrained models publicly available at: https://github.com/chi0tzp/WarpedGANSpace.

This paper is on video recognition using Transformers. Very recent attempts in this area have demonstrated promising results in terms of recognition accuracy, yet they have been also shown to induce, in many cases, significant computational overheads due to the additional modelling of the temporal information. In this work, we propose a Video Transformer model the complexity of which scales linearly with the number of frames in the video sequence and hence induces no overhead compared to an image-based Transformer model. To achieve this, our model makes two approximations to the full space-time attention used in Video Transformers: (a) It restricts time attention to a local temporal window and capitalizes on the Transformer's depth to obtain full temporal coverage of the video sequence. (b) It uses efficient space-time mixing to attend jointly spatial and temporal locations without inducing any additional cost on top of a spatial-only attention model. We also show how to integrate 2 very lightweight mechanisms for global temporal-only attention which provide additional accuracy improvements at minimal computational cost. We demonstrate that our model produces very high recognition accuracy on the most popular video recognition datasets while at the same time being significantly more efficient than other Video Transformer models.

The prominence of deep learning, large amount of annotated data and increasingly powerful hardware made it possible to reach remarkable performance for supervised classification tasks, in many cases saturating the training sets. However the resulting models are specialized to a single very specific task and domain. Adapting the learned classification to new domains is a hard problem due to at least three reasons: (1) the new domains and the tasks might be drastically different; (2) there might be very limited amount of annotated data on the new domain and (3) full training of a new model for each new task is prohibitive in terms of computation and memory, due to the sheer number of parameters of deep CNNs. In this paper, we present a method to learn new-domains and tasks incrementally, building on prior knowledge from already learned tasks and without catastrophic forgetting. We do so by jointly parametrizing weights across layers using low-rank Tucker structure. The core is task agnostic while a set of task specific factors are learnt on each new domain. We show that leveraging tensor structure enables better performance than simply using matrix operations. Joint tensor modelling also naturally leverages correlations across different layers. Compared with previous methods which have focused on adapting each layer separately, our approach results in more compact representations for each new task/domain. We apply the proposed method to the 10 datasets of the Visual Decathlon Challenge and show that our method offers on average about 7.5× reduction in number of parameters and competitive performance in terms of both classification accuracy and Decathlon score.

It is known that facial landmarks provide pose, expression and shape information. In addition, when matching, for example, a profile and/or expressive face to a frontal one, knowledge of these landmarks is useful for establishing correspondence which can help improve recognition. However, in prior work on face recognition, facial landmarks are only used for face cropping in order to remove scale, rotation and translation variations. This paper proposes a simple approach to face recognition which gradually integrates features from different layers of a facial landmark localization network into different layers of the recognition network. To this end, we propose an appropriate feature integration layer which makes the features compatible before integration. We show that such a simple approach systematically improves recognition on the most difficult face recognition datasets, setting a new state-of-the-art on IJB-B, IJB-C and MegaFace datasets.

Several studies show that animal needs are often expressed through their faces. Though remarkable progress has been made towards the automatic understanding of human faces, this has not been the case with animal faces. There exists significant room for algorithmic advances that could realize automatic systems for interpreting animal faces. Besides scientific value, resulting technology will foster better and cheaper animal care. We believe the underlying research progress is mainly obstructed by the lack of an adequately annotated dataset of animal faces, covering a wide spectrum of animal species. To this end, we introduce a large-scale, hierarchical annotated dataset of animal faces, featuring 22.4K faces from 350 diverse species and 21 animal orders across biological taxonomy. These faces are captured `in-the-wild' conditions and are consistently annotated with 9 landmarks on key facial features. The dataset is structured and scalable by design; its development underwent four systematic stages involving rigorous, overall effort of over 6K man-hours. We benchmark it for face alignment using the existing art under two new problem settings. Results showcase its challenging nature, unique attributes and present definite prospects for novel, adaptive, and generalized face-oriented CV algorithms. Further benchmarking the dataset across face detection and fine-grained recognition tasks demonstrates its multi-task applications and room for improvement. The dataset is available at: https://fdmaproject.wordpress.com/.

This paper is on highly accurate and highly efficient human pose estimation. Recent works based on Fully Convolutional Networks (FCNs) have demonstrated excellent results for this difficult problem. While residual connections within FCNs have proved to be quintessential for achieving high accuracy, we re-analyze this design choice in the context of improving both the accuracy and the efficiency over the state-of- the-art. In particular, we make the following contributions: (a) We propose gated skip connections with per-channel learnable parameters to control the data flow for each channel within the module within the macro-module. (b) We introduce a hybrid network that combines the HourGlass and U-Net architectures which minimizes the number of identity connections within the network and increases the performance for the same parameter budget. Our model achieves state-of-the-art results on the MPII and LSP datasets. In addition, with a reduction of 3× in model size and complexity, we show no decrease in performance when compared to the original HourGlass network.

Lip-reading models have been significantly improved recently thanks to powerful deep learning architectures. However, most works focused on frontal or near frontal views of the mouth. As a consequence, lip-reading performance seriously deteriorates in non-frontal mouth views. In this work, we present a framework for training pose-invariant lip-reading models on synthetic data instead of collecting and annotating non-frontal data which is costly and tedious. The proposed model significantly outperforms previous approaches on non-frontal views while retaining the superior performance on frontal and near frontal mouth views. Specifically, we propose to use a 3D Morphable Model (3DMM) to augment LRW, an existing large-scale but mostly frontal dataset, by generating synthetic facial data in arbitrary poses. The newly derived dataset, is used to train a state-of-the-art neural network for lip-reading. We conducted a cross-database experiment for isolated word recognition on the LRS2 dataset, and reported an absolute improvement of 2.55%. The benefit of the proposed approach becomes clearer in extreme poses where an absolute improvement of up to 20.64% over the baseline is achieved.

We present the first direct approach targeted explicitly on human hands that infers depth from monocular RGB images. We achieve this with a Convolutional Neural Network (CNN) that employs a stacked hourglass model as its main building block. Intermediate supervision is used in several outputs of the proposed architecture in a staged approach. To aid the process of training and inference, hand segmentation masks are also estimated in such intermediate supervision steps, and used to guide the subsequent depth estimation process. In order to train and evaluate the proposed method we compile and make publicly available HandRGBD, a new dataset of 20,601 views of hands, each consisting of an RGB image and an aligned depth map. Based on HandRGBD, we explore variants of the proposed approach in an ablative study and determine the most accurate one. The results of an extensive experimental evaluation demonstrate that hand depth estimation from a single RGB frame can be achieved with an accuracy of 22mm, which is comparable to the accuracy achieved by contemporary low-cost depth cameras. Such a 3D reconstruction of hands based on RGB information is valuable as a final result on its own right, but also as an input to several other hand analysis and perception algorithms that require depth input. In this context, the proposed approach bridges the gap between RGB and RGBD, by making all existing RGBD-based methods applicable to RGB input.

This paper addresses the problem of unsupervised discovery of object landmarks. We take a different path compared to that of existing works, based on 2 novel perspectives: (1) Self-training: starting from generic keypoints, we propose a self-training approach where the goal is to learn a detector that improves itself becoming more and more tuned to object landmarks. (2) Correspondence: we identify correspondence as a key objective for unsupervised landmark discovery and propose an optimization scheme which alternates between recovering object landmark correspondence across different images via clustering and learning an object landmark descriptor without labels. Compared to previous works, our approach can learn landmarks that are more flexible in terms of capturing large changes in viewpoint. We show the favourable properties of our method on a variety of difficult datasets including LS3D, BBCPose and Human3.6M. Code is available at https://github.com/malldimi1/UnsupervisedLandmarks

Our goal is to design architectures that retain the groundbreaking performance of Convolutional Neural Networks (CNNs) for landmark localization and at the same time are lightweight, compact and suitable for applications with limited computational resources. To this end, we make the following contributions: (a) we are the first to study the effect of neural network binarization on localization tasks, namely human pose estimation and face alignment. We exhaustively evaluate various design choices, identify performance bottlenecks, and more importantly propose multiple orthogonal ways to boost performance. (b) Based on our analysis, we propose a novel hierarchical, parallel and multi-scale residual architecture that yields large performance improvement over the standard bottleneck block while having the same number of parameters, thus bridging the gap between the original network and its binarized counterpart. (c) We perform a large number of ablation studies that shed light on the properties and the performance of the proposed block. (d) We present results for experiments on the most challenging datasets for human pose estimation and face alignment, reporting in many cases state-of-the-art performance. (e) We further provide additional results for the problem of facial part segmentation. Code can be downloaded from https://www.adrianbulat.com/binary-cnn-landmarks.

Recent findings indicate that over-parametrization, while crucial for successfully training deep neural networks, also introduces large amounts of redundancy. Tensor methods have the potential to efficiently parametrize over-complete representations by leveraging this redundancy. In this paper, we propose to fully parametrize Convolutional Neural Networks (CNNs) with a single high-order, low-rank tensor. Previous works on network tensorization have focused on parametrizing individual layers (convolutional or fully connected) only, and perform the tensorization layer-by-layer separately. In contrast, we propose to jointly capture the full structure of a neural network by parametrizing it with a single high-order tensor, the modes of which represent each of the architectural design parameters of the network (e.g. number of convolutional blocks, depth, number of stacks, input features, etc). This parametrization allows to regularize the whole network and drastically reduce the number of parameters. Our model is end-to-end trainable and the low-rank structure imposed on the weight tensor acts as an implicit regularization. We study the case of networks with rich structure, namely Fully Convolutional Networks (FCNs), which we propose to parametrize with a single 8th-order tensor. We show that our approach can achieve superior performance with small compression rates, and attain high compression rates with negligible drop in accuracy for the challenging task of human pose estimation.

This paper proposes a method to ease the unsupervised learning of object landmark detectors. Similarly to previous methods, our approach is fully unsupervised in a sense that it does not require or make any use of annotated landmarks for the target object category. Contrary to previous works, we do however assume that a landmark detector, which has already learned a structured representation for a given object category in a fully supervised manner, is available. Under this setting, our main idea boils down to adapting the given pre-trained network to the target object categories in a fully unsupervised manner. To this end, our method uses the pre-trained network as a core which remains frozen and does not get updated during training, and learns, in an unsupervised manner, only a projection matrix to perform the adaptation to the target categories. By building upon an existing structured representation learned in a supervised manner, the optimization problem solved by our method is much more constrained with significantly less parameters to learn which seems to be important for the case of unsupervised learning. We show that our method surpasses fully unsupervised techniques trained from scratch as well as a strong baseline based on fine-tuning, and produces state-of-the-art results on several datasets. Code can be found at tiny.cc/GitHub-Unsupervised

This paper addresses 2 challenging tasks: improving the quality of low resolution facial images and accurately locating the facial landmarks on such poor resolution images. To this end, we make the following 5 contributions: (a) we propose Super-FAN: the very first end-to-end system that addresses both tasks simultaneously, i.e. both improves face resolution and detects the facial landmarks. The novelty or Super-FAN lies in incorporating structural information in a GAN-based super-resolution algorithm via integrating a sub-network for face alignment through heatmap regression and optimizing a novel heatmap loss. (b) We illustrate the benefit of training the two networks jointly by reporting good results not only on frontal images (as in prior work) but on the whole spectrum of facial poses, and not only on synthetic low resolution images (as in prior work) but also on real-world images. (c) We improve upon the state-of-the-art in face super-resolution by proposing a new residual-based architecture. (d) Quantitatively, we show large improvement over the state-of-the-art for both face super-resolution and alignment. (e) Qualitatively, we show for the first time good results on real-world low resolution images like the ones of Fig. 1.

Several end-to-end deep learning approaches have been recently presented which extract either audio or visual features from the input images or audio signals and perform speech recognition. However, research on end-to-end audiovisual models is very limited. In this work, we present an end-to-end audiovisual model based on residual networks and Bidirectional Gated Recurrent Units (BGRUs). To the best of our knowledge, this is the first audiovisual fusion model which simultaneously learns to extract features directly from the image pixels and audio waveforms and performs within-context word recognition on a large publicly available dataset (LRW). The model consists of two streams, one for each modality, which extract features directly from mouth regions and raw waveforms. The temporal dynamics in each stream/modality are modeled by a 2-layer BGRU and the fusion of multiple streams/modalities takes place via another 2-layer BGRU. A slight improvement in the classification rate over an end-to-end audio-only and MFCC-based model is reported in clean audio conditions and low levels of noise. In presence of high levels of noise, the end-to-end audiovisual model significantly outperforms both audio-only models.

In this paper we present a deep learning architecture for extracting word embeddings for visual speech recognition. The embeddings summarize the information of the mouth region that is relevant to the problem of word recognition, while suppressing other types of variability such as speaker, pose and illumination. The system is comprised of a spatiotemporal convolutional layer, a Residual Network and bidirectional LSTMs and is trained on the Lipreading in-the-wild database. We first show that the proposed architecture goes beyond state-of-the-art on closed-set word identification, by attaining 11.92% error rate on a vocabulary of 500 words. We then examine the capacity of the embeddings in modelling words unseen during training. We deploy Probabilistic Linear Discriminant Analysis (PLDA) to model the embeddings and perform low-shot learning experiments on words unseen during training. The experiments demonstrate that word-level visual speech recognition is feasible even in cases where the target words are not included in the training set.

Recent works in speech recognition rely either on connectionist temporal classification (CTC) or sequence-to-sequence models for character-level recognition. CTC assumes conditional independence of individual characters, whereas attention-based models can provide nonsequential alignments. Therefore, we could use a CTC loss in combination with an attention-based model in order to force monotonic alignments and at the same time get rid of the conditional independence assumption. In this paper, we use the recently proposed hybrid CTC/attention architecture for audio-visual recognition of speech in-the-wild. To the best of our knowledge, this is the first time that such a hybrid architecture architecture is used for audio-visual recognition of speech. We use the LRS2 database and show that the proposed audio-visual model leads to an 1.3% absolute decrease in word error rate over the audio-only model and achieves the new state-of-the-art performance on LRS2 database (7% word error rate). We also observe that the audio-visual model significantly outperforms the audio-based model (up to 32.9% absolute improvement in word error rate) for several different types of noise as the signal-to-noise ratio decreases.

The papers in this special section examine the concept of automated face analysis (AFA). AFA has received special attention from the computer vision and pattern recognition communities. Research progress often gives the impression that problems such as face recognition and face detection are solved, at least for some scenarios. Several aspects of face analysis remain open problems, including the implementation of large scale face recognition/detection methods for in the wild images, emotion recognition, micro-expression analysis, and others. The community keeps making rapid progress on these topics, with continual improvement of current methods and creation of new ones that push the state-of-the-art. Applications are countless, including security and video surveillance, human computer/robot interaction, communication, entertainment, and commerce, while having an important social impact in assistive technologies for education and health. The importance of face analysis, together with the vast amount of work on the subject and the latest developments in the field, motivated us to organize a special section on this theme. The scope of the compilation comprises all aspects of face analysis from a computer vision perspective. Including, but not limited to: recognition, detection, alignment, reconstruction of faces, pose estimation of faces, gaze analysis, age, emotion, gender, and facial attributes estimation, and applications among others.

Linear regression is a fundamental building block in many face detection and tracking algorithms, typically used to predict shape displacements from image features through a linear mapping. This paper presents a Functional Regression solution to the least squares problem, which we coin Continuous Regression, resulting in the first real-time incremental face tracker. Contrary to prior work in Functional Regression, in which B-splines or Fourier series were used, we propose to approximate the input space by its first-order Taylor expansion, yielding a closed-form solution for the continuous domain of displacements. We then extend the continuous least squares problem to correlated variables, and demonstrate the generalisation of our approach. We incorporate Continuous Regression into the cascaded regression framework, and show its computational benefits for both training and testing. We then present a fast approach for incremental learning within Cascaded Continuous Regression, coined iCCR, and show that its complexity allows real-time face tracking, being 20 times faster than the state of the art. To the best of our knowledge, this is the first incremental face tracker that is shown to operate in real-time. We show that iCCR achieves state-of-the-art performance on the 300-VW dataset, the most recent, large-scale benchmark for face tracking.

This paper investigates how far a very deep neural network is from attaining close to saturating performance on existing 2D and 3D face alignment datasets. To this end, we make the following 5 contributions: (a) we construct, for the first time, a very strong baseline by combining a state-of-the-art architecture for landmark localization with a state-of-the-art residual block, train it on a very large yet synthetically expanded 2D facial landmark dataset and finally evaluate it on all other 2D facial landmark datasets. (b)We create a guided by 2D landmarks network which converts 2D landmark annotations to 3D and unifies all existing datasets, leading to the creation of LS3D-W, the largest and most challenging 3D facial landmark dataset to date (~230,000 images). (c) Following that, we train a neural network for 3D face alignment and evaluate it on the newly introduced LS3D-W. (d) We further look into the effect of all “traditional” factors affecting face alignment performance like large pose, initialization and resolution, and introduce a “new” one, namely the size of the network. (e) We show that both 2D and 3D face alignment networks achieve performance of remarkable accuracy which is probably close to saturating the datasets used. Training and testing code as well as the dataset can be downloaded from https://www.adrianbulat.com/face-alignment/.

Our goal is to design architectures that retain the groundbreaking performance of CNNs for landmark localization and at the same time are lightweight, compact and suitable for applications with limited computational resources. To this end, we make the following contributions: (a) we are the first to study the effect of neural network binarization on localization tasks, namely human pose estimation and face alignment. We exhaustively evaluate various design choices, identify performance bottlenecks, and more importantly propose multiple orthogonal ways to boost performance. (b) Based on our analysis, we propose a novel hierarchical, parallel and multi-scale residual architecture that yields large performance improvement over the standard bottleneck block while having the same number of parameters, thus bridging the gap between the original network and its binarized counterpart. (c) We perform a large number of ablation studies that shed light on the properties and the performance of the proposed block. (d) We present results for experiments on the most challenging datasets for human pose estimation and face alignment, reporting in many cases state-of-the-art performance. Code can be downloaded from https://www.adrianbulat.com/binary-cnn-landmarks.

3D face reconstruction is a fundamental Computer Vision problem of extraordinary difficulty. Current systems often assume the availability of multiple facial images (sometimes from the same subject) as input, and must address a number of methodological challenges such as establishing dense correspondences across large facial poses, expressions, and non-uniform illumination. In general these methods require complex and inefficient pipelines for model building and fitting. In this work, we propose to address many of these limitations by training a Convolutional Neural Network (CNN) on an appropriate dataset consisting of 2D images and 3D facial models or scans. Our CNN works with just a single 2D facial image, does not require accurate alignment nor establishes dense correspondence between images, works for arbitrary facial poses and expressions, and can be used to reconstruct the whole 3D facial geometry (including the non-visible parts of the face) bypassing the construction (during training) and fitting (during testing) of a 3D Morphable Model. We achieve this via a simple CNN architecture that performs direct regression of a volumetric representation of the 3D facial geometry from a single 2D image. We also demonstrate how the related task of facial landmark localization can be incorporated into the proposed framework and help improve reconstruction quality, especially for the cases of large poses and facial expressions. Code and models will be made available at http://aaronsplace.co.uk.

An open question in facial landmark localization in video is whether one should perform tracking or tracking-by-detection (i.e. face alignment). Tracking produces fittings of high accuracy but is prone to drifting. Tracking-by-detection is drift-free but results in low accuracy fittings. To provide a solution to this problem, we describe the very first, to the best of our knowledge, synergistic approach between detection (face alignment) and tracking which completely eliminates drifting from face tracking, and does not merely perform tracking-by-detection. Our first main contribution is to show that one can achieve this synergy between detection and tracking using a principled optimization framework based on the theory of Global Variable Consensus Optimization using ADMM; Our second contribution is to show how the proposed analytic framework can be integrated within state-of-the-art discriminative methods for face alignment and tracking based on cascaded regression and deeply learned features. Overall, we call our method Discriminative Global Consensus Model (DGCM). Our third contribution is to show that DGCM achieves large performance improvement over the currently best performing face tracking methods on the most challenging category of the 300-VW dataset.

18th Annual Conference of the International Speech Communication Association, INTERSPEECH 2017, Stockholm, Sweden, 20-24 August 2017

We propose an end-to-end deep learning architecture for word-level visual speech recognition. The system is a combination of spatiotemporal convolutional, residual and bidirectional Long Short-Term Memory networks. We train and evaluate it on the Lipreading In-The-Wild benchmark, a challenging database of 500-size target-words consisting of 1.28sec video excerpts from BBC TV broadcasts. The proposed network attains word accuracy equal to 83.0, yielding 6.8 absolute improvement over the current state-of-the-art, without using information about word boundaries during training or testing.

Fitting algorithms for Active Appearance Models (AAMs) are usually considered to be robust but slow or fast but less able to generalize well to unseen variations. In this paper, we look into AAM fitting algorithms and make the following orthogonal contributions: We present a simple "project-out" optimization framework that unifies and revises the most well-known optimization problems and solutions in AAMs. Based on this framework, we describe robust simultaneous AAM fitting algorithms the complexity of which is not prohibitive for current systems. We then go on one step further and propose a new approximate project-out AAM fitting algorithm which we coin Extended Project-Out Inverse Compositional (E-POIC). In contrast to current algorithms, E-POIC is both efficient and robust. Next, we describe a part-based AAM employing a translational motion model, which results in superior fitting and convergence properties. We also show that the proposed AAMs, when trained "in-the-wild" using SIFT descriptors, perform surprisingly well even for the case of unseen unconstrained images. Via a number of experiments on unconstrained human and animal face databases, we show that our combined contributions largely bridge the gap between exact and current approximate methods for AAM fitting and perform comparably with state-of-the-art face alignment systems.

Active appearance models (AAMs) are generative models of shape and appearance that have proven very attractive for their ability to handle wide changes in illumination, pose, and occlusion when trained in the wild, while not requiring large training data set like regression-based or deep learning methods. The problem of fitting an AAM is usually formulated as a non-linear least squares one and the main way of solving it is a standard Gauss-Newton algorithm. In this paper, we extend AAMs in two ways: we first extend the Gauss-Newton framework by formulating a bidirectional fitting method that deforms both the image and the template to fit a new instance. We then formulate a second order method by deriving an efficient Newton method for AAMs fitting. We derive both methods in a unified framework for two types of AAMs, holistic and part-based, and additionally show how to exploit the structure in the problem to derive fast yet exact solutions. We perform a thorough evaluation of all algorithms on three challenging and recently annotated in-the-wild data sets, and investigate fitting accuracy, convergence properties, and the influence of noise in the initialization. We compare our proposed methods to other algorithms and show that they yield state-of-the-art results, out-performing other methods while having superior convergence properties.

In this work, the focus is on gradient-based correlation schemes which constitute an alternative to phase correlation for FFT-based motion estimation. In particular, our contribution is threefold. First, we present an analysis which highlights the key features of gradient schemes. Second, we introduce an illumination invariant to gradient correlation. Third, we provide a comparison of gradient schemes in the application of block matching for video coding and draw several useful observations and conclusions.

Image registration under challenging realistic conditions is a very important area of research. In this paper, we focus on algorithms that seek to densely align two volumetric images according to a global similarity measure. Despite intensive research in this area, there is still a need for similarity measures that are robust to outliers common to many different types of images. For example, medical image data is often corrupted by intensity inhomogeneities and may contain outliers in the form of pathologies. In this paper we propose a global similarity measure that is robust to both intensity inhomogeneities and outliers without requiring prior knowledge of the type of outliers. We combine the normalised gradients of images with the cosine function and show that it is theoretically robust against a very general class of outliers. Experimentally, we verify the robustness of our measures within two distinct algorithms. Firstly, we embed our similarity measures within a proof-of-concept extension of the Lucas–Kanade algorithm for volumetric data. Finally, we embed our measures within a popular non-rigid alignment framework based on free-form deformations and show it to be robust against both simulated tumours and intensity inhomogeneities.

In this paper we evaluate the notion of scene analysis with regard to risk. We consider the problem of evaluating risk and potential hazards in an environment and providing a quantified risk score. A definition of risk is given incorporating two elements, Firstly scene stability, where Newtonian Physics are introduced into the scene analysis process, evaluating object stability within a scene. The effectiveness of which is demonstrated by conducting experiments on several scenes including a variety of stability levels. Secondly the analysis of the intrinsic risk related properties of an object, which is estimated using learning techniques and the utilisation of the 3D Voxel HOG descriptor, analysed against the state-of-the-art descriptors. Finally a new dataset is provided that is designed for scene analysis focusing on risk evaluation.

Cascaded regression approaches have been recently shown to achieve state-of-the-art performance for many computer vision tasks. Beyond its connection to boosting, cascaded regression has been interpreted as a learning-based approach to iterative optimization methods like the Newton's method. However, in prior work, the connection to optimization theory is limited only in learning a mapping from image features to problem parameters. In this paper, we consider the problem of facial deformable model fitting using cascaded regression and make the following contributions: (a) We propose regression to learn a sequence of averaged Jacobian and Hessian matrices from data, and from them descent directions in a fashion inspired by Gauss-Newton optimization. (b) We show that the optimization problem in hand has structure and devise a learning strategy for a cascaded regression approach that takes the problem structure into account. By doing so, the proposed method learns and employs a sequence of averaged Jacobians and descent directions in a subspace orthogonal to the facial appearance variation; hence, we call it Project-Out Cascaded Regression (PO-CR). (c) Based on the principles of PO-CR, we built a face alignment system that produces remarkably accurate results on the challenging iBUG data set outperforming previously proposed systems by a large margin. Code for our system is available from http://www.cs.nott.ac.uk/~yzt/.

Finding landmarks on objects like faces is a challenging computer vision problem, especially in real life conditions (or in-the-wild) and Active Appearance Models have been widely used to solve it. State-of-the-art algorithms for fitting an AAM to a new image are based on Gauss-Newton (GN) optimization. Recently fast GN algorithms have been proposed for both forward additive and inverse compositional fitting frameworks. In this paper, we propose a fast and exact bi-directional (Fast-Bd) approach to AAM fitting by combining both approaches. Although such a method might appear to increase computational burden, we show that by capitalizing on results from optimization theory, an exact solution, as computationally efficient as the original forward or inverse formulation, can be derived. Our proposed bi-directional approach achieves state-of-the-art performance and superior convergence properties. These findings are validated on two challenging, in-the-wild data sets, LFPW and Helen, and comparison is provided to the state-of-the art methods for Active Appearance Models fitting.

We propose a face alignment framework that relies on the texture model generated by the responses of discriminatively trained part-based filters. Unlike standard texture models built from pixel intensities or responses generated by generic filters (e.g. Gabor), our framework has two important advantages. First, by virtue of discriminative training, invariance to external variations (like identity, pose, illumination and expression) is achieved. Second, we show that the responses generated by discriminatively trained filters (or patch-experts) are sparse and can be modeled using a very small number of parameters. As a result, the optimization methods based on the proposed texture model can better cope with unseen variations. We illustrate this point by formulating both part-based and holistic approaches for generic face alignment and show that our framework outperforms the state-of-the-art on multiple "wild" databases. The code and dataset annotations are available for research purposes from http://ibug.doc.ic.ac.uk/resources.

Lucas-Kanade and active appearance models are among the most commonly used methods for image alignment and facial fitting, respectively. They both utilize nonlinear gradient descent, which is usually applied on intensity values. In this paper, we propose the employment of highly descriptive, densely sampled image features for both problems. We show that the strategy of warping the multichannel dense feature image at each iteration is more beneficial than extracting features after warping the intensity image at each iteration. Motivated by this observation, we demonstrate robust and accurate alignment and fitting performance using a variety of powerful feature descriptors. Especially with the employment of histograms of oriented gradient and scale-invariant feature transform features, our method significantly outperforms the current state-of-the-art results on in-the-wild databases.

Arguably, Deformable Part Models (DPMs) are one of the most prominent approaches for face alignment with impressive results being recently reported for both controlled lab and unconstrained settings. Fitting in most DPM methods is typically formulated as a two-step process during which discriminatively trained part templates are first correlated with the image to yield a filter response for each landmark and then shape optimization is performed over these filter responses. This process, although computationally efficient, is based on fixed part templates which are assumed to be independent, and has been shown to result in imperfect filter responses and detection ambiguities. To address this limitation, in this paper, we propose to jointly optimize a part-based, trained in-the-wild, flexible appearance model along with a global shape model which results in a joint translational motion model for the model parts via Gauss-Newton (GN) optimization. We show how significant computational reductions can be achieved by building a full model during training but then efficiently optimizing the proposed cost function on a sparse grid using weighted least-squares during fitting. We coin the proposed formulation Gauss-Newton Deformable Part Model (GN-DPM). Finally, we compare its performance against the state-of-the-art and show that the proposed GN-DPM outperforms it, in some cases, by a large margin. Code for our method is available from http://ibug.doc.ic.ac.uk/resources.

We propose the combination of dense Histogram of Oriented Gradients (HOG) features with Active Appearance Models (AAMs). We employ the efficient Inverse Compositional optimization technique and show results for the task of face fitting. By taking advantage of the descriptive characteristics of HOG features, we build robust and accurate AAMs that generalize well to unseen faces with illumination, identity, pose and occlusion variations. Our experiments on challenging in-the-wild databases show that HOG AAMs significantly outperfrom current state-of-the-art results of discriminative methods trained on larger databases.

Active Appearance Models (AAMs) are statistical models of shape and appearance widely used in computer vision to detect landmarks on objects like faces. Fitting an AAM to a new image can be formulated as a non-linear least-squares problem which is typically solved using iterative methods. Owing to its efficiency, Gauss-Newton optimization has been the standard choice over more sophisticated approaches like Newton. In this paper, we show that the AAM problem has structure which can be used to solve efficiently the original Newton problem without any approximations. We then make connections to the original Gauss-Newton algorithm and study experimentally the effect of the additional terms introduced by the Newton formulation on both fitting accuracy and convergence. Based on our derivations, we also propose a combined Newton and Gauss-Newton method which achieves promising fitting and convergence performance. Our findings are validated on two challenging in-the-wild data sets.

We present Active Orientation Models (AOMs), generative models of facial shape and appearance, which extend the well-known paradigm of Active Appearance Models (AAMs) for the case of generic face alignment under unconstrained conditions. Robustness stems from the fact that the proposed AOMs employ a statistically robust appearance model based on the principal components of image gradient orientations. We show that when incorporated within standard optimization frameworks for AAM learning and fitting, this kernel Principal Component Analysis results in robust algorithms for model fitting. At the same time, the resulting optimization problems maintain the same computational cost. As a result, the main similarity of AOMs with AAMs is the computational complexity. In particular, the project-out version of AOMs is as computationally efficient as the standard project-out inverse compositional algorithm, which is admittedly one of the fastest algorithms for fitting AAMs. We verify experimentally that: 1) AOMs generalize well to unseen variations and 2) outperform all other state-of-the-art AAM methods considered by a large margin. This performance improvement brings AOMs at least in par with other contemporary methods for face alignment. Finally, we provide MATLAB code at http://ibug.doc.ic.ac.uk/resources.

We describe a very simple framework for deriving the most-well known optimization problems in Active Appearance Models (AAMs), and most importantly for providing efficient solutions. Our formulation results in two optimization problems for fast and exact AAM fitting, and one new algorithm which has the important advantage of being applicable to 3D. We show that the dominant cost for both forward and inverse algorithms is a few times mN which is the cost of projecting an image onto the appearance subspace. This makes both algorithms not only computationally realizable but also very attractive speed-wise for most current systems. Because exact AAM fitting is no longer computationally prohibitive, we trained AAMs in-the-wild with the goal of investigating whether AAMs benefit from such a training process. Our results show that although we did not use sophisticated shape priors, robust features or robust norms for improving performance, AAMs perform notably well and in some cases comparably with current state-of-the-art methods. We provide Matlab source code for training, fitting and reproducing the results presented in this paper at http://ibug.doc.ic.ac.uk/resources.

Principal Component Analysis (PCA) is perhaps the most prominent learning tool for dimensionality reduction in pattern recognition and computer vision. However, the ℓ 2-norm employed by standard PCA is not robust to outliers. In this paper, we propose a kernel PCA method for fast and robust PCA, which we call Euler-PCA (e-PCA). In particular, our algorithm utilizes a robust dissimilarity measure based on the Euler representation of complex numbers. We show that Euler-PCA retains PCA’s desirable properties while suppressing outliers. Moreover, we formulate Euler-PCA in an incremental learning framework which allows for efficient computation. In our experiments we apply Euler-PCA to three different computer vision applications for which our method performs comparably with other state-of-the-art approaches.

We introduce the notion of subspace learning from image gradient orientations for appearance-based object recognition. As image data are typically noisy and noise is substantially different from Gaussian, traditional subspace learning from pixel intensities very often fails to estimate reliably the low-dimensional subspace of a given data population. We show that replacing pixel intensities with gradient orientations and the ℓ₂ norm with a cosine-based distance measure offers, to some extend, a remedy to this problem. Within this framework, which we coin Image Gradient Orientations (IGO) subspace learning, we first formulate and study the properties of Principal Component Analysis of image gradient orientations (IGO-PCA). We then show its connection to previously proposed robust PCA techniques both theoretically and experimentally. Finally, we derive a number of other popular subspace learning techniques, namely, Linear Discriminant Analysis (LDA), Locally Linear Embedding (LLE), and Laplacian Eigenmaps (LE). Experimental results show that our algorithms significantly outperform popular methods such as Gabor features and Local Binary Patterns and achieve state-of-the-art performance for difficult problems such as illumination and occlusion-robust face recognition. In addition to this, the proposed IGO-methods require the eigendecomposition of simple covariance matrices and are as computationally efficient as their corresponding ℓ₂ norm intensity-based counterparts. Matlab code for the methods presented in this paper can be found at http://ibug.doc.ic.ac.uk/resources.

We propose an exact framework for online learning with a family of indefinite (not positive) kernels. As we study the case of nonpositive kernels, we first show how to extend kernel principal component analysis (KPCA) from a reproducing kernel Hilbert space to Krein space. We then formulate an incremental KPCA in Krein space that does not require the calculation of preimages and therefore is both efficient and exact. Our approach has been motivated by the application of visual tracking for which we wish to employ a robust gradient-based kernel. We use the proposed nonlinear appearance model learned online via KPCA in Krein space for visual tracking in many popular and difficult tracking scenarios. We also show applications of our kernel framework for the problem of face recognition.

We propose a robust approach to discriminant kernel-based feature extraction for face recognition and verification. We show, for the first time, how to perform the eigen analysis of the within-class scatter matrix directly in the feature space. This eigen analysis provides the eigenspectrum of its range space and the corresponding eigenvectors as well as the eigenvectors spanning its null space. Based on our analysis, we propose a kernel discriminant analysis (KDA) which combines eigenspectrum regularization with a feature-level scheme (ER-KDA). Finally, we combine the proposed ER-KDA with a nonlinear robust kernel particularly suitable for face recognition/verification applications which require robustness against outliers caused by occlusions and illumination changes. We applied the proposed framework to several popular databases (Yale, AR, XM2VTS) and achieved state-of-the-art performance for most of our experiments.

We introduce a fast and robust subspace-based approach to appearance-based object tracking. The core of our approach is based on Fast Robust Correlation (FRC), a recently proposed technique for the robust estimation of large translational displacements. We show how the basic principles of FRC can be naturally extended to formulate a robust version of Principal Component Analysis (PCA) which can be efficiently implemented incrementally and therefore is particularly suitable for robust real-time appearance-based object tracking. Our experimental results demonstrate that the proposed approach outperforms other state-of-the-art holistic appearance-based trackers on several popular video sequences.

We introduce the notion of Principal Component Analysis (PCA) of image gradient orientations. As image data is typically noisy, but noise is substantially different from Gaussian, traditional PCA of pixel intensities very often fails to estimate reliably the low-dimensional subspace of a given data population. We show that replacing intensities with gradient orientations and the ℓ2norm with a cosine-based distance measure offers, to some extend, a remedy to this problem. Our scheme requires the eigen-decomposition of a covariance matrix and is as computationally efficient as standard ℓ2intensity-based PCA. We demonstrate some of its favorable properties for the application of face recognition.

We investigate classification of non-linguistic vocalisations with a novel audiovisual approach and Long Short-Term Memory (LSTM) Recurrent Neural Networks as highly successful dynamic sequence classifiers. As database of evaluation serves this year's Paralinguistic Challenge's Audiovisual Interest Corpus of human-to-human natural conversation. For video-based analysis we compare shape and appearance based features. These are fused in an early manner with typical audio descriptors. The results show significant improvements of LSTM networks over a static approach based on Support Vector Machines. More important, we can show a significant gain in performance when fusing audio and visual shape features.

We propose a correlation-based approach to parametric object alignment particularly suitable for face analysis applications which require efficiency and robustness against occlusions and illumination changes. Our algorithm registers two images by iteratively maximizing their correlation coefficient using gradient ascent. We compute this correlation coefficient from complex gradients which capture the orientation of image structures rather than pixel intensities. The maximization of this gradient correlation coefficient results in an algorithm which is as computationally efficient as ℓ2norm-based algorithms, can be extended within the inverse compositional framework (without the need for Hessian recomputation) and is robust to outliers. To the best of our knowledge, no other algorithm has been proposed so far having all three features. We show the robustness of our algorithm for the problem of face alignment in the presence of occlusions and non-uniform illumination changes. The code that reproduces the results of our paper can be found at http://ibug.doc.ic.ac.uk/resources.

We address the problem of subpixel registration of images assumed to be related by a pure translation. We present a method which extends gradient correlation to achieve subpixel accuracy. Our scheme is based on modeling the dominant singular vectors of the 2-D gradient correlation matrix with a generic kernel which we derive by studying the structure of gradient correlation assuming natural image statistics. Our kernel has a parametric form which offers flexibility in modeling the functions obtained from various types of image data. We estimate the kernel parameters, including the unknown subpixel shifts, using the Levenberg-Marquardt algorithm. Experiments with LANDSAT and MRI data show that our scheme outperforms recently proposed state-of-the-art phase correlation methods.

In this paper, we present a new method for bidirectional relighting for 3D-aided 2D face recognition under large pose and illumination changes. During subject enrollment, we build subject-specific 3D annotated models by using the subjects' raw 3D data and 2D texture. During authentication, the probe 2D images are projected onto a normalized image space using the subject-specific 3D model in the gallery. Then, a bidirectional relighting algorithm and two similarity metrics (a view-dependent complex wavelet structural similarity and a global similarity) are employed to compare the gallery and probe. We tested our algorithms on the UHDB11 and UHDB12 databases that contain 3D data with probe images under large lighting and pose variations. The experimental results show the robustness of our approach in recognizing faces in difficult situations.

In this paper we examine a new penalty term for the Bayesian Information Criterion (BIC) that is suited to the problem of speaker diarization. Based on our previous approach of penalizing each cluster only with its effective sample size - an approach we called segmental - we propose a stricter penalty term. The criterion we derive retains the main property of the Segmental-BIC, i.e. it approximates the evidence of overall partitions of the data and simultaneously leads to a pairwise dissimilarity measure that is completely defined by the pair of clusters in question. The experimental results show significant improvement in diarization accuracy on the ESTER benchmark.

We present a robust FFT-based approach to scale-invariant image registration. Our method relies on FFT-based correlation twice: once in the log-polar Fourier domain to estimate the scaling and rotation and once in the spatial domain to recover the residual translation. Previous methods based on the same principles are not robust. To equip our scheme with robustness and accuracy, we introduce modifications which tailor the method to the nature of images. First, we derive efficient log-polar Fourier representations by replacing image functions with complex gray-level edge maps. We show that this representation both captures the structure of salient image features and circumvents problems related to the low-pass nature of images, interpolation errors, border effects, and aliasing. Second, to recover the unknown parameters, we introduce the normalized gradient correlation. We show that, using image gradients to perform correlation, the errors induced by outliers are mapped to a uniform distribution for which our normalized gradient correlation features robust performance. Exhaustive experimentation with real images showed that, unlike any other Fourier-based correlation techniques, the proposed method was able to estimate translations, arbitrary rotations, and scale factors up to 6.

A fast and robust gradient-based motion estimation technique which operates in the frequency domain is presented. The algorithm combines the natural advantages of a good feature selection offered by gradient-based methods with the robustness and speed provided by FFT-based correlation schemes. Experimentation with real images taken from a popular database showed that, unlike any other Fourier-based techniques, the method was able to estimate translations, arbitrary rotations and scale factors in the range 4-6.

In this paper, the problem of moment-based shape orientation and symmetry classification is jointly considered. A generalization and modification of current state-of-the-art geometric moment-based functions is introduced. The properties of these functions are investigated thoroughly using Fourier series analysis and several observations and closed-form solutions are derived. We demonstrate the connection between the results presented in this work and symmetry detection principles suggested from previous complex moment-based formulations. The proposed analysis offers a unifying framework for shape orientation/symmetry detection. In the context of symmetry classification and matching, the second part of this work presents a frequency domain method, aiming at computing a robust moment-based feature set based on a true polar Fourier representation of image complex gradients and a novel periodicity detection scheme using subspace analysis. The proposed approach removes the requirement for accurate shape centroid estimation, which is the main limitation of moment-based methods, operating in the image spatial domain. The proposed framework demonstrated improved performance, compared to state-of-the-art methods.

A novel frequency domain approach to roto-translation estimation is presented. The baseline gradient cross-correlation method is extended to handle rotations. A key feature of the proposed scheme is the ability to achieve good performance in the presence of both large translations and rotations as well as noise, a scenario for which other Fourier-based methods typically fail. Robustness and accuracy in conjunction with computational efficiency, offered by the frequency domain formulation, make the algorithm useful in a number of image processing tasks such as image registration. 1

A frequency domain approach for the detection of symmetries in real images is presented. Our framework is based on recent state-of-the-art research where motion estimation techniques are employed to sequentially determine all the associated parameters. In particular, we introduce several modifications regarding the order of symmetry estimation and the detection of the axes of possible bilateral symmetry. Preliminary results demonstrate the efficiency of our approach.

The problem of shape-based recognition of objects under affine transformations is considered. We focus on the construction of a robust and highly discriminative affine invariant function that can be used for within-class object recognition applications. Using the boundaries of the objects of interest, a training scheme, based on principal component analysis (PCA), is proposed to derive a set of basis functions with desired properties. The derived bases are then used for the construction of a novel affine invariant function. The proposed invariant function is evaluated for the problem of aircraft silhouette identification and appears to achieve comparable performance to a popular wavelet-based affine invariant function. At the same time, the proposed framework is much simpler than that based on wavelet analysis.

A scheme, based on principal component analysis (PCA), is proposed that can be used for the recognition of 2-D planar shapes under affine transformations. A PCA step is first used to map the object boundary to its canonical form, reducing the problem of the nonuniform sampling of the object contour introduced by the affine transformation. Then, a PCA-based scheme is employed to train a set of basis functions on the signals extracted from the objects' boundaries. The derived bases are used to analyze the boundary locally. Based on the theory of invariants and local boundary analysis, a novel invariant function is constructed. The performance of the proposed framework is compared with a standard wavelet-based approach with promising results.