Real-time Memory Efficient Large-pose Face Alignment via Deep Evolutionary Network

There is an urgent need to apply face alignment in a memory-efficient and real-time manner due to the recent explosion of face recognition applications. However, impact factors such as large pose variation and computational inefficiency, still hinder its broad implementation. To this end, we propose a computationally efficient deep evolutionary model integrated with 3D Diffusion Heap Maps (DHM). First, we introduce a sparse 3D DHM to assist the initial modeling process under extreme pose conditions. Afterward, a simple and effective CNN feature is extracted and fed to Recurrent Neural Network (RNN) for evolutionary learning. To accelerate the model, we propose an efficient network structure to accelerate the evolutionary learning process through a factorization strategy. Extensive experiments on three popular alignment databases demonstrate the advantage of the proposed models over the state-of-the-art, especially under large-pose conditions. Notably, the computational speed of our model is 10 times faster than the state-of-the-art on CPU and 14 times on GPU. We also discuss and analyze the limitations of our models and future research work.

Generative One-Shot Face Recognition

One-shot face recognition measures the ability to identify persons with only seeing them at one glance, and is a hallmark of human visual intelligence. It is challenging for conventional machine learning approaches to mimic this way, since limited data are hard to effectively represent the data variance. The goal of one-shot face recognition is to learn a large-scale face recognizer, which is capable to fight off the data imbalance challenge. In this paper, we propose a novel generative adversarial one-shot face recognizer, attempting to synthesize meaningful data for one-shot classes by adapting the data variances from other normal classes. Specifically, we target at building a more effective general face classifier for both normal persons and one-shot persons. Technically, we design a new loss function by formulating knowledge transfer generator and a general classifier into a unified framework. Such a two-player minimax optimization can guide the generation of more effective data, which effectively promote the underrepresented classes in the learned model and lead to a remarkable improvement in face recognition performance. We evaluate our proposed model on the MS-Celeb-1M one-shot learning benchmark task, where we could recognize 94.98% of the test images at the precision of 99% for the one-shot classes, keeping an overall Top1 accuracy at $99.80\%$ for the normal classes. To the best of our knowledge, this is the best performance among all the published methods using this benchmark task with the same setup, including all the participants in the recent MS-Celeb-1M challenge at ICCV 2017\footnote{http://www.msceleb.org/challenge2/2017}.

Rethinking Zero-Shot Learning: A Conditional Visual Classification Perspective

Zero-shot learning (ZSL) aims to recognize instances of unseen classes solely based on the semantic descriptions of the classes. Existing algorithms usually formulate it as a semantic-visual correspondence problem, by learning mappings from one feature space to the other. Despite being reasonable, previous approaches essentially discard the highly precious discriminative power of visual features in an implicit way, and thus produce undesirable results. We instead reformulate ZSL as a conditioned visual classification problem, i.e., classifying visual features based on the classifiers learned from the semantic descriptions. With this reformulation, we develop algorithms targeting various ZSL settings: For the conventional setting, we propose to train a deep neural network that directly generates visual feature classifiers from the semantic attributes with an episode-based training scheme; For the generalized setting, we concatenate the learned highly discriminative classifiers for seen classes and the generated classifiers for unseen classes to classify visual features of all classes; For the transductive setting, we exploit unlabeled data to effectively calibrate the classifier generator using a novel learning-without-forgetting self-training mechanism and guide the process by a robust generalized cross-entropy loss. Extensive experiments show that our proposed algorithms significantly outperform state-of-the-art methods by large margins on most benchmark datasets in all the ZSL settings.

Visual Semantic Reasoning for Image-Text Matching

Image-text matching has been a hot research topic bridging the vision and language areas. It remains challenging because the current representation of image usually lacks global semantic concepts as in its corresponding text caption. To address this issue, we propose a simple and interpretable reasoning model to generate visual representation that captures key objects and semantic concepts of a scene. Specifically, we first build up connections between image regions and perform reasoning with Graph Convolutional Networks to generate features with semantic relationships. Then, we propose to use the gate and memory mechanism to perform global semantic reasoning on these relationship-enhanced features, select the discriminative information and gradually generate the representation for the whole scene. Experiments validate that our method achieves a new state-of-the-art for the image-text matching on MS-COCO and Flickr30K datasets. It outperforms the current best method by 6.8% relatively for image retrieval and 4.8% relatively for caption retrieval on MS-COCO (Recall@1 using 1K test set). On Flickr30K, our model improves image retrieval by 12.6% relatively and caption retrieval by 5.8% relatively (Recall@1). Our code is available at https://github.com/KunpengLi1994/VSRN.

Double-Head RCNN: Rethinking Classification and Localization for Object Detection

Modern R-CNN based detectors apply a head to extract Region of Interest (RoI) features for both classification and localization tasks. In contrast, we found that these two tasks have opposite preferences towards two widely used head structures (i.e. fully connected head and convolution head). Specifically, the fully connected head is more suitable for the classification task, while the convolution head is more suitable for the localization task. Therefore, we propose a Double-Head method, which has a fully connected head focusing on classification and a convolution head to pay more attention to bounding box regression. In addition, we have two findings for the unfocused tasks (i.e. classification in the convolution head, and bounding box regression in the fully connected head): (a) adding classification to the convolution head is complementary to the classification in the fully connected head, and (b) bounding box regression provides auxiliary supervision for the fully connected head. Without bells and whistles, our method gains +3.5 and +2.8 AP on MS COCO dataset from Feature Pyramid Network (FPN) baselines with ResNet-50 and ResNet-101 backbones, respectively.

Large Scale Incremental Learning

Modern machine learning suffers from catastrophic forgetting when learning new classes incrementally. The performance dramatically degrades due to the missing data of old classes. Incremental learning methods have been proposed to retain the knowledge acquired from the old classes, by using knowledge distilling and keeping a few exemplars from the old classes. However, these methods struggle to scale up to a large number of classes. We believe this is because of the combination of two factors: (a) the data imbalance between the old and new classes, and (b) the increasing number of visually similar classes. Distinguishing between an increasing number of visually similar classes is particularly challenging, when the training data is unbalanced. We propose a simple and effective method to address this data imbalance issue. We found that the last fully connected layer has a strong bias towards the new classes, and this bias can be corrected by a linear model. With two bias parameters, our method performs remarkably well on two large datasets: ImageNet (1000 classes) and MS-Celeb-1M (10000 classes), outperforming the state-of-the-art algorithms by 11.1% and 13.2% respectively.

Multimodal Style Transfer via Graph Cuts

An assumption widely used in recent neural style transfer methods is that image styles can be described by global statics of deep features like Gram or covariance matrices. Alternative approaches have represented styles by decomposing them into local pixel or neural patches. Despite the recent progress, most existing methods treat the semantic patterns of style image uniformly, resulting unpleasing results on complex styles. In this paper, we introduce a more flexible and general universal style transfer technique: multimodal style transfer (MST). MST explicitly considers the matching of semantic patterns in content and style images. Specifically, the style image features are clustered into sub-style components, which are matched with local content features under a graph cut formulation. A reconstruction network is trained to transfer each sub-style and render the final stylized result. Extensive experiments demonstrate the superior effectiveness, robustness and flexibility of MST.

Spatially Constrained Generative Adversarial Networks for Conditional Image Generation

Image generation has raised tremendous attention in both academic and industrial areas, especially for the conditional and target-oriented image generation, such as criminal portrait and fashion design. Although the current studies have achieved preliminary results along this direction, they always focus on class labels as the condition where spatial contents are randomly generated from latent vectors. Edge details are usually blurred since spatial information is difficult to preserve. In light of this, we propose a novel Spatially Constrained Generative Adversarial Network (SCGAN), which decouples the spatial constraints from the latent vector and makes these constraints feasible as additional controllable signals. To enhance the spatial controllability, a generator network is specially designed to take a semantic segmentation, a latent vector and an attribute-level label as inputs step by step. Besides, a segmentor network is constructed to impose spatial constraints on the generator. Experimentally, we provide both visual and quantitative results on CelebA and DeepFashion datasets, and demonstrate that the proposed SCGAN is very effective in controlling the spatial contents as well as generating high-quality images.

EV-Action: Electromyography-Vision Multi-Modal Action Dataset

Multi-modal human motion analysis is a critical and attractive research topic. Most existing multi-modal action datasets only provide visual modalities such as RGB, depth, or low quality skeleton data. In this paper, we introduce a new, large-scale dataset named EV-Action dataset. It consists RGB, depth, electromyography (EMG), and two skeleton modalities. Compared with others, our dataset has two major improvements: (1) we deploy a motion capturing system to obtain high quality skeleton modality, which provides more comprehensive motion information including skeleton, trajectory, and acceleration with higher accuracy, sample frequency, and more skeleton markers. (2) we include an EMG modality. While EMG is used as an effective indicator for biomechanics area, it has yet to be well explored in multimedia, computer vision, and machine learning areas. To the best of our knowledge, this is the first action dataset with EMG modality. In this paper, we introduce the details of EV-Action dataset. A simple yet effective framework for EMG-based action recognition is proposed. Moreover, we provide state-of-the-art baselines for each modality. The approaches achieve considerable improvements when EMG is involved, and it demonstrates the effectiveness of EMG modality in human action analysis tasks. We hope this dataset could make significant contributions to signal processing, multimedia, computer vision, machine learning, biomechanics, and other interdisciplinary fields.

Rethinking Classification and Localization in R-CNN

Modern R-CNN based detectors share the RoI feature extractor head for both classification and localization tasks, based upon the correlation between the two tasks. In contrast, we found that different head structures (i.e. fully connected head and convolution head) have opposite preferences towards these two tasks. Specifically, the fully connected head is more suitable for the classification task, while the convolution head is more suitable for the localization task. Therefore, we propose a double-head method to separate these two tasks into different heads (i.e. a fully connected head for classification and a convolution head for box regression). Without bells and whistles, our method gains +3.4 and +2.7 points mAP on MS COCO dataset from Feature Pyramid Network baselines with ResNet-50 and ResNet-101 backbones, respectively.