1.On the ''steerability" of generative adversarial networks ⬇️
An open secret in contemporary machine learning is that many models work beautifully on standard benchmarks but fail to generalize outside the lab. This has been attributed to training on biased data, which provide poor coverage over real world events. Generative models are no exception, but recent advances in generative adversarial networks (GANs) suggest otherwise -- these models can now synthesize strikingly realistic and diverse images. Is generative modeling of photos a solved problem? We show that although current GANs can fit standard datasets very well, they still fall short of being comprehensive models of the visual manifold. In particular, we study their ability to fit simple transformations such as camera movements and color changes. We find that the models reflect the biases of the datasets on which they are trained (e.g., centered objects), but that they also exhibit some capacity for generalization: by "steering" in latent space, we can shift the distribution while still creating realistic images. We hypothesize that the degree of distributional shift is related to the breadth of the training data distribution, and conduct experiments that demonstrate this. Code is released on our project page: this https URL
2.Predicting Next-Season Designs on High Fashion Runway ⬇️
Fashion is a large and fast-changing industry. Foreseeing the upcoming fashion trends is beneficial for fashion designers, consumers, and retailers. However, fashion trends are often perceived as unpredictable due to the enormous amount of factors involved into designers' subjectivity. In this paper, we propose a fashion trend prediction framework and design neural network models to leverage structured fashion runway show data, learn the fashion collection embedding, and further train RNN/LSTM models to capture the designers' style evolution. Our proposed framework consists of (1) a runway embedding learning model that uses fashion runway images to learn every season's collection embedding, and (2) a next-season fashion design prediction model that leverage the concept of designer style and trend to predict next-season design given designers. Through experiments on a collected dataset across 32 years of fashion shows, our framework can achieve the best performance of 78.42% AUC on average and 95% for an individual designer when predicting the next season's design.
3.EnforceNet: Monocular Camera Localization in Large Scale Indoor Sparse LiDAR Point Cloud ⬇️
Pose estimation is a fundamental building block for robotic applications such as autonomous vehicles, UAV, and large scale augmented reality. It is also a prohibitive factor for those applications to be in mass production, since the state-of-the-art, centimeter-level pose estimation often requires long mapping procedures and expensive localization sensors, e.g. LiDAR and high precision GPS/IMU, etc. To overcome the cost barrier, we propose a neural network based solution to localize a consumer degree RGB camera within a prior sparse LiDAR map with comparable centimeter-level precision. We achieved it by introducing a novel network module, which we call resistor module, to enforce the network generalize better, predicts more accurately, and converge faster. Such results are benchmarked by several datasets we collected in the large scale indoor parking garage scenes. We plan to open both the data and the code for the community to join the effort to advance this field.
4.Efficient Segmentation: Learning Downsampling Near Semantic Boundaries ⬇️
Many automated processes such as auto-piloting rely on a good semantic segmentation as a critical component. To speed up performance, it is common to downsample the input frame. However, this comes at the cost of missed small objects and reduced accuracy at semantic boundaries. To address this problem, we propose a new content-adaptive downsampling technique that learns to favor sampling locations near semantic boundaries of target classes. Cost-performance analysis shows that our method consistently outperforms the uniform sampling improving balance between accuracy and computational efficiency. Our adaptive sampling gives segmentation with better quality of boundaries and more reliable support for smaller-size objects.
5.How much real data do we actually need: Analyzing object detection performance using synthetic and real data ⬇️
In recent years, deep learning models have resulted in a huge amount of progress in various areas, including computer vision. By nature, the supervised training of deep models requires a large amount of data to be available. This ideal case is usually not tractable as the data annotation is a tremendously exhausting and costly task to perform. An alternative is to use synthetic data. In this paper, we take a comprehensive look into the effects of replacing real data with synthetic data. We further analyze the effects of having a limited amount of real data. We use multiple synthetic and real datasets along with a simulation tool to create large amounts of cheaply annotated synthetic data. We analyze the domain similarity of each of these datasets. We provide insights about designing a methodological procedure for training deep networks using these datasets.
6.Pedestrian Tracking by Probabilistic Data Association and Correspondence Embeddings ⬇️
This paper studies the interplay between kinematics (position and velocity) and appearance cues for establishing correspondences in multi-target pedestrian tracking. We investigate tracking-by-detection approaches based on a deep learning detector, joint integrated probabilistic data association (JIPDA), and appearance-based tracking of deep correspondence embeddings. We first addressed the fixed-camera setup by fine-tuning a convolutional detector for accurate pedestrian detection and combining it with kinematic-only JIPDA. The resulting submission ranked first on the 3DMOT2015 benchmark. However, in sequences with a moving camera and unknown ego-motion, we achieved the best results by replacing kinematic cues with global nearest neighbor tracking of deep correspondence embeddings. We trained the embeddings by fine-tuning features from the second block of ResNet-18 using angular loss extended by a margin term. We note that integrating deep correspondence embeddings directly in JIPDA did not bring significant improvement. It appears that geometry of deep correspondence embeddings for soft data association needs further investigation in order to obtain the best from both worlds.
7.Uncertainty-aware Self-ensembling Model for Semi-supervised 3D Left Atrium Segmentation ⬇️
Training deep convolutional neural networks usually requires a large amount of labeled data. However, it is expensive and time-consuming to annotate data for medical image segmentation tasks. In this paper, we present a novel uncertainty-aware semi-supervised framework for left atrium segmentation from 3D MR images. Our framework can effectively leverage the unlabeled data by encouraging consistent predictions of the same input under different perturbations. Concretely, the framework consists of a student model and a teacher model, and the student model learns from the teacher model by minimizing a segmentation loss and a consistency loss with respect to the targets of the teacher model. We design a novel uncertainty-aware scheme to enable the student model to gradually learn from the meaningful and reliable targets by exploiting the uncertainty information. Experiments show that our method achieves high performance gains by incorporating the unlabeled data. Our method outperforms the state-of-the-art semi-supervised methods, demonstrating the potential of our framework for the challenging semi-supervised problems.
8.Data Selection for training Semantic Segmentation CNNs with cross-dataset weak supervision ⬇️
Training convolutional networks for semantic segmentation with strong (per-pixel) and weak (per-bounding-box) supervision requires a large amount of weakly labeled data. We propose two methods for selecting the most relevant data with weak supervision. The first method is designed for finding visually similar images without the need of labels and is based on modeling image representations with a Gaussian Mixture Model (GMM). As a byproduct of GMM modeling, we present useful insights on characterizing the data generating distribution. The second method aims at finding images with high object diversity and requires only the bounding box labels. Both methods are developed in the context of automated driving and experimentation is conducted on Cityscapes and Open Images datasets. We demonstrate performance gains by reducing the amount of employed weakly labeled images up to 100 times for Open Images and up to 20 times for Cityscapes.
9.Improving Semantic Segmentation via Dilated Affinity ⬇️
Introducing explicit constraints on the structural predictions has been an effective way to improve the performance of semantic segmentation models. Existing methods are mainly based on insufficient hand-crafted rules that only partially capture the image structure, and some methods can also suffer from the efficiency issue. As a result, most of the state-of-the-art fully convolutional networks did not adopt these techniques. In this work, we propose a simple, fast yet effective method that exploits structural information through direct supervision with minor additional expense. To be specific, our method explicitly requires the network to predict semantic segmentation as well as dilated affinity, which is a sparse version of pair-wise pixel affinity. The capability of telling the relationships between pixels are directly built into the model and enhance the quality of segmentation in two stages. 1) Joint training with dilated affinity can provide robust feature representations and thus lead to finer segmentation results. 2) The extra output of affinity information can be further utilized to refine the original segmentation with a fast propagation process. Consistent improvements are observed on various benchmark datasets when applying our framework to the existing state-of-the-art model. Codes will be released soon.
10.Perception of visual numerosity in humans and machines ⬇️
Numerosity perception is foundational to mathematical learning, but its computational bases are strongly debated. Some investigators argue that humans are endowed with a specialized system supporting numerical representation; others argue that visual numerosity is estimated using continuous magnitudes, such as density or area, which usually co-vary with number. Here we reconcile these contrasting perspectives by testing deep networks on the same numerosity comparison task that was administered to humans, using a stimulus space that allows to measure the contribution of non-numerical features. Our model accurately simulated the psychophysics of numerosity perception and the associated developmental changes: discrimination was driven by numerosity information, but non-numerical features had a significant impact, especially early during development. Representational similarity analysis further highlighted that both numerosity and continuous magnitudes were spontaneously encoded even when no task had to be carried out, demonstrating that numerosity is a major, salient property of our visual environment.
11.Speed estimation evaluation on the KITTI benchmark based on motion and monocular depth information ⬇️
In this technical report we investigate speed estimation of the ego-vehicle on the KITTI benchmark using state-of-the-art deep neural network based optical flow and single-view depth prediction methods. Using a straightforward intuitive approach and approximating a single scale factor, we evaluate several application schemes of the deep networks and formulate meaningful conclusions such as: combining depth information with optical flow improves speed estimation accuracy as opposed to using optical flow alone; the quality of the deep neural network methods influences speed estimation performance; using the depth and optical flow results from smaller crops of wide images degrades performance. With these observations in mind, we achieve a RMSE of less than 1 m/s for vehicle speed estimation using monocular images as input from recordings of the KITTI benchmark. Limitations and possible future directions are discussed as well.
12.A Short Note on the Kinetics-700 Human Action Dataset ⬇️
We describe an extension of the DeepMind Kinetics human action dataset from 600 classes to 700 classes, where for each class there are at least 600 video clips from different YouTube videos. This paper details the changes introduced for this new release of the dataset, and includes a comprehensive set of statistics as well as baseline results using the I3D neural network architecture.
13.A Unified Deep Framework for Joint 3D Pose Estimation and Action Recognition from a Single RGB Camera ⬇️
We present a deep learning-based multitask framework for joint 3D human pose estimation and action recognition from RGB video sequences. Our approach proceeds along two stages. In the first, we run a real-time 2D pose detector to determine the precise pixel location of important keypoints of the body. A two-stream neural network is then designed and trained to map detected 2D keypoints into 3D poses. In the second, we deploy the Efficient Neural Architecture Search (ENAS) algorithm to find an optimal network architecture that is used for modeling the spatio-temporal evolution of the estimated 3D poses via an image-based intermediate representation and performing action recognition. Experiments on Human3.6M, MSR Action3D and SBU Kinect Interaction datasets verify the effectiveness of the proposed method on the targeted tasks. Moreover, we show that our method requires a low computational budget for training and inference.
14.Fused Detection of Retinal Biomarkers in OCT Volumes ⬇️
Optical Coherence Tomography (OCT) is the primary imaging modality for detecting pathological biomarkers associated to retinal diseases such as Age-Related Macular Degeneration. In practice, clinical diagnosis and treatment strategies are closely linked to biomarkers visible in OCT volumes and the ability to identify these plays an important role in the development of ophthalmic pharmaceutical products. In this context, we present a method that automatically predicts the presence of biomarkers in OCT cross-sections by incorporating information from the entire volume. We do so by adding a bidirectional LSTM to fuse the outputs of a Convolutional Neural Network that predicts individual biomarkers. We thus avoid the need to use pixel-wise annotations to train our method, and instead provide fine-grained biomarker information regardless. On a dataset of 416 volumes, we show that our approach imposes coherence between biomarker predictions across volume slices and our predictions are superior to several existing approaches.
15.Semi-supervised Breast Lesion Detection in Ultrasound Video Based on Temporal Coherence ⬇️
Breast lesion detection in ultrasound video is critical for computer-aided diagnosis. However, detecting lesion in video is quite challenging due to the blurred lesion boundary, high similarity to soft tissue and lack of video annotations. In this paper, we propose a semi-supervised breast lesion detection method based on temporal coherence which can detect the lesion more accurately. We aggregate features extracted from the historical key frames with adaptive key-frame scheduling strategy. Our proposed method accomplishes the unlabeled videos detection task by leveraging the supervision information from a different set of labeled images. In addition, a new WarpNet is designed to replace both the traditional spatial warping and feature aggregation operation, leading to a tremendous increase in speed. Experiments on 1,060 2D ultrasound sequences demonstrate that our proposed method achieves state-of-the-art video detection result as 91.3% in mean average precision and 19 ms per frame on GPU, compared to a RetinaNet based detection method in 86.6% and 32 ms.
16.Human Pose Estimation for Real-World Crowded Scenarios ⬇️
Human pose estimation has recently made significant progress with the adoption of deep convolutional neural networks. Its many applications have attracted tremendous interest in recent years. However, many practical applications require pose estimation for human crowds, which still is a rarely addressed problem. In this work, we explore methods to optimize pose estimation for human crowds, focusing on challenges introduced with dense crowds, such as occlusions, people in close proximity to each other, and partial visibility of people. In order to address these challenges, we evaluate three aspects of a pose detection approach: i) a data augmentation method to introduce robustness to occlusions, ii) the explicit detection of occluded body parts, and iii) the use of the synthetic generated datasets. The first approach to improve the accuracy in crowded scenarios is to generate occlusions at training time using person and object cutouts from the object recognition dataset COCO (Common Objects in Context). Furthermore, the synthetically generated dataset JTA (Joint Track Auto) is evaluated for the use in real-world crowd applications. In order to overcome the transfer gap of JTA originating from a low pose variety and less dense crowds, an extension dataset is created to ease the use for real-world applications. Additionally, the occlusion flags provided with JTA are utilized to train a model, which explicitly distinguishes between occluded and visible body parts in two distinct branches. The combination of the proposed additions to the baseline method help to improve the overall accuracy by 4.7% AP and thereby provide comparable results to current state-of-the-art approaches on the respective dataset.
17.Mango Tree Net -- A fully convolutional network for semantic segmentation and individual crown detection of mango trees ⬇️
This work presents a method for semantic segmentation of mango trees in high resolution aerial imagery, and, a novel method for individual crown detection of mango trees using segmentation output. Mango Tree Net, a fully convolutional neural network (FCN), is trained using supervised learning to perform semantic segmentation of mango trees in imagery acquired using an unmanned aerial vehicle (UAV). The proposed network is retrained to separate touching/overlapping tree crowns in segmentation output. Contour based connected object detection is performed on the segmentation output from retrained network. Bounding boxes are drawn on the original images using coordinates of connected objects to achieve individual crown detection. The training dataset consists of 8,824 image patches of size 240 x 240. The approach is tested for performance on segmentation and individual crown detection tasks using test datasets containing 36 and 4 images respectively. The performance is analyzed using standard metrics precision, recall, f1-score and accuracy. Results obtained demonstrate the robustness of the proposed methods despite variations in factors such as scale, occlusion, lighting conditions and surrounding vegetation.
18.A General Framework for Uncertainty Estimation in Deep Learning ⬇️
End-to-end learning has recently emerged as a promising technique to tackle the problem of autonomous driving. Existing works show that learning a navigation policy from raw sensor data may reduce the system's reliance on external sensing systems, (e.g. GPS), and/or outperform traditional methods based on state estimation and planning. However, existing end-to-end methods generally trade off performance for safety, hindering their diffusion to real-life applications. For example, when confronted with an input which is radically different from the training data, end-to-end autonomous driving systems are likely to fail, compromising the safety of the vehicle. To detect such failure cases, this work proposes a general framework for uncertainty estimation which enables a policy trained end-to-end to predict not only action commands, but also a confidence about its own predictions. In contrast to previous works, our framework can be applied to any existing neural network and task, without the need to change the network's architecture or loss, or to train the network. In order to do so, we generate confidence levels by forward propagation of input and model uncertainties using Bayesian inference. We test our framework on the task of steering angle regression for an autonomous car, and compare our approach to existing methods with both qualitative and quantitative results on a real dataset. Finally, we show an interesting by-product of our framework: robustness against adversarial attacks.
19.Learning Depth from Monocular Videos Using Synthetic Data: A Temporally-Consistent Domain Adaptation Approach ⬇️
Majority of state-of-the-art monocular depth estimation methods are supervised learning approaches. The success of such approaches heavily depends on the high-quality depth labels which are expensive to obtain. Recent methods try to learn depth networks by exploring unsupervised cues from monocular videos which are easier to acquire but less reliable. In this paper, we propose to resolve this dilemma by transferring knowledge from synthetic videos with easily obtainable ground truth depth labels. Due to the stylish difference between synthetic and real images, we propose a temporally-consistent domain adaptation (TCDA) approach that simultaneously explores labels in the synthetic domain and temporal constraints in the videos to improve style transfer and depth prediction. Furthermore, we make use of the ground truth optical flow and pose information in the synthetic data to learn moving mask and pose prediction networks. The learned moving masks can filter out moving regions that produces erroneous temporal constraints and the estimated poses provide better initializations for estimating temporal constraints. The experimental results demonstrate the effectiveness of our method and comparable performance against state-of-the-art.
20.Cascade RetinaNet: Maintaining Consistency for Single-Stage Object Detection ⬇️
Recent researches attempt to improve the detection performance by adopting the idea of cascade for single-stage detectors. In this paper, we analyze and discover that inconsistency is the major factor limiting the performance. The refined anchors are associated with the feature extracted from the previous location and the classifier is confused by misaligned classification and localization. Further, we point out two main designing rules for the cascade manner: improving consistency between classification confidence and localization performance, and maintaining feature consistency between different stages. A multistage object detector named Cas-RetinaNet, is then proposed for reducing the misalignments. It consists of sequential stages trained with increasing IoU thresholds for improving the correlation, and a novel Feature Consistency Module for mitigating the feature inconsistency. Experiments show that our proposed Cas-RetinaNet achieves stable performance gains across different models and input scales. Specifically, our method improves RetinaNet from 39.1 AP to 41.1 AP on the challenging MS COCO dataset without any bells or whistles.
21.Separable Convolutional LSTMs for Faster Video Segmentation ⬇️
Semantic Segmentation is an important module for autonomous robots such as self-driving cars. The advantage of video segmentation approaches compared to single image segmentation is that temporal image information is considered, and their performance increases due to this. Hence, single image segmentation approaches are extended by recurrent units such as convolutional LSTM (convLSTM) cells, which are placed at suitable positions in the basic network architecture. However, a major critique of video segmentation approaches based on recurrent neural networks is their large parameter count and their computational complexity, and so, their inference time of one video frame takes up to 66 percent longer than their basic version. Inspired by the success of the spatial and depthwise separable convolutional neural networks, we generalize these techniques for convLSTMs in this work, so that the number of parameters and the required FLOPs are reduced significantly. Experiments on different datasets show that the segmentation approaches using the proposed, modified convLSTM cells achieve similar or slightly worse accuracy, but are up to 15 percent faster on a GPU than the ones using the standard convLSTM cells. Furthermore, a new evaluation metric is introduced, which measures the amount of flickering pixels in the segmented video sequence.
22.Deep inspection: an electrical distribution pole parts study via deep neural networks ⬇️
Electrical distribution poles are important assets in electricity supply. These poles need to be maintained in good condition to ensure they protect community safety, maintain reliability of supply, and meet legislative obligations. However, maintaining such a large volumes of assets is an expensive and challenging task. To address this, recent approaches utilise imagery data captured from helicopter and/or drone inspections. Whilst reducing the cost for manual inspection, manual analysis on each image is still required. As such, several image-based automated inspection systems have been proposed. In this paper, we target two major challenges: tiny object detection and extremely imbalanced datasets, which currently hinder the wide deployment of the automatic inspection. We propose a novel two-stage zoom-in detection method to gradually focus on the object of interest. To address the imbalanced dataset problem, we propose the resampling as well as reweighting schemes to iteratively adapt the model to the large intra-class variation of major class and balance the contributions to the loss from each class. Finally, we integrate these components together and devise a novel automatic inspection framework. Extensive experiments demonstrate that our proposed approaches are effective and can boost the performance compared to the baseline methods.
23.Stereo-based terrain traversability analysis using normal-based segmentation and superpixel surface analysis ⬇️
In this paper, an stereo-based traversability analysis approach for all terrains in off-road mobile robotics, e.g. Unmanned Ground Vehicles (UGVs) is proposed. This approach reformulates the problem of terrain traversability analysis into two main problems: (1) 3D terrain reconstruction and (2) terrain all surfaces detection and analysis. The proposed approach is using stereo camera for perception and 3D reconstruction of the terrain. In order to detect all the existing surfaces in the 3D reconstructed terrain as superpixel surfaces (i.e. segments), an image segmentation technique is applied using geometry-based features (pixel-based surface normals). Having detected all the surfaces, Superpixel Surface Traversability Analysis approach (SSTA) is applied on all of the detected surfaces (superpixel segments) in order to classify them based on their traversability index. The proposed SSTA approach is based on: (1) Superpixel surface normal and plane estimation, (2) Traversability analysis using superpixel surface planes. Having analyzed all the superpixel surfaces based on their traversability, these surfaces are finally classified into five main categories as following: traversable, semi-traversable, non-traversable, unknown and undecided.
24.Instant Motion Tracking and Its Applications to Augmented Reality ⬇️
Augmented Reality (AR) brings immersive experiences to users. With recent advances in computer vision and mobile computing, AR has scaled across platforms, and has increased adoption in major products. One of the key challenges in enabling AR features is proper anchoring of the virtual content to the real world, a process referred to as tracking. In this paper, we present a system for motion tracking, which is capable of robustly tracking planar targets and performing relative-scale 6DoF tracking without calibration. Our system runs in real-time on mobile phones and has been deployed in multiple major products on hundreds of millions of devices.
25.2nd Place Solution to the GQA Challenge 2019 ⬇️
We present a simple method that achieves unexpectedly superior performance for Complex Reasoning involved Visual Question Answering. Our solution collects statistical features from high-frequency words of all the questions asked about an image and use them as accurate knowledge for answering further questions of the same image. We are fully aware that this setting is not ubiquitously applicable, and in a more common setting one should assume the questions are asked separately and they cannot be gathered to obtain a knowledge base. Nonetheless, we use this method as an evidence to demonstrate our observation that the bottleneck effect is more severe on the feature extraction part than it is on the knowledge reasoning part. We show significant gaps when using the same reasoning model with 1) ground-truth features; 2) statistical features; 3) detected features from completely learned detectors, and analyze what these gaps mean to researches on visual reasoning topics. Our model with the statistical features achieves the 2nd place in the GQA Challenge 2019.
26.Rethinking RGB-D Salient Object Detection: Models, Datasets, and Large-Scale Benchmarks ⬇️
The use of RGB-D information for salient object detection has been explored in recent years. However, relatively few efforts have been spent in modeling salient object detection over real-world human activity scenes with RGB-D. In this work, we fill the gap by making the following contributions to RGB-D salient object detection. First, we carefully collect a new salient person (SIP) dataset, which consists of 1K high-resolution images that cover diverse real-world scenes from various viewpoints, poses, occlusion, illumination, and background. Second, we conduct a large-scale and so far the most comprehensive benchmark comparing contemporary methods, which has long been missing in the area and can serve as a baseline for future research. We systematically summarized 31 popular models, evaluated 17 state-of-the-art methods over seven datasets with totally about 91K images. Third, we propose a simple baseline architecture, called Deep Depth-Depurator Network (D3Net). It consists of a depth depurator unit and a feature learning module, performing initial low-quality depth map filtering and cross-modal feature learning respectively. These components form a nested structure and are elaborately designed to be learned jointly. D3Net exceeds the performance of any prior contenders across five metrics considered, thus serves as a strong baseline to advance the research frontier. We also demonstrate that D3Net can be used to efficiently extract salient person masks from the real scenes, enabling effective background changed book cover application with 20 fps on a single GPU. All the saliency maps, our new SIP dataset, baseline model, and evaluation tools are made publicly available at this https URL.
27.Improving 3D Object Detection for Pedestrians with Virtual Multi-View Synthesis Orientation Estimation ⬇️
Accurately estimating the orientation of pedestrians is an important and challenging task for autonomous driving because this information is essential for tracking and predicting pedestrian behavior. This paper presents a flexible Virtual Multi-View Synthesis module that can be adopted into 3D object detection methods to improve orientation estimation. The module uses a multi-step process to acquire the fine-grained semantic information required for accurate orientation estimation. First, the scene's point cloud is densified using a structure preserving depth completion algorithm and each point is colorized using its corresponding RGB pixel. Next, virtual cameras are placed around each object in the densified point cloud to generate novel viewpoints, which preserve the object's appearance. We show that this module greatly improves the orientation estimation on the challenging pedestrian class on the KITTI benchmark. When used with the open-source 3D detector AVOD-FPN, we outperform all other published methods on the pedestrian Orientation, 3D, and Bird's Eye View benchmarks.
28.Efficient Pipeline for Camera Trap Image Review ⬇️
Biologists all over the world use camera traps to monitor biodiversity and wildlife population density. The computer vision community has been making strides towards automating the species classification challenge in camera traps, but it has proven difficult to to apply models trained in one region to images collected in different geographic areas. In some cases, accuracy falls off catastrophically in new region, due to both changes in background and the presence of previously-unseen species. We propose a pipeline that takes advantage of a pre-trained general animal detector and a smaller set of labeled images to train a classification model that can efficiently achieve accurate results in a new region.
29.AugLabel: Exploiting Word Representations to Augment Labels for Face Attribute Classification ⬇️
Augmenting data in image space (eg. flipping, cropping etc) and activation space (eg. dropout) are being widely used to regularise deep neural networks and have been successfully applied on several computer vision tasks. Unlike previous works, which are mostly focused on doing augmentation in the aforementioned domains, we propose to do augmentation in label space. In this paper, we present a novel method to generate fixed dimensional labels with continuous values for images by exploiting the word2vec representations of the existing categorical labels. We then append these representations with existing categorical labels and train the model. We validated our idea on two challenging face attribute classification data sets viz. CelebA and LFWA. Our extensive experiments show that the augmented labels improve the performance of the competitive deep learning baseline and reduce the need of annotated real data up to 50%, while attaining a performance similar to the state-of-the-art methods.
30.Real-time Hair Segmentation and Recoloring on Mobile GPUs ⬇️
We present a novel approach for neural network-based hair segmentation from a single camera input specifically designed for real-time, mobile application. Our relatively small neural network produces a high-quality hair segmentation mask that is well suited for AR effects, e.g. virtual hair recoloring. The proposed model achieves real-time inference speed on mobile GPUs (30-100+ FPS, depending on the device) with high accuracy. We also propose a very realistic hair recoloring scheme. Our method has been deployed in major AR application and is used by millions of users.
31.Real-time Facial Surface Geometry from Monocular Video on Mobile GPUs ⬇️
We present an end-to-end neural network-based model for inferring an approximate 3D mesh representation of a human face from single camera input for AR applications. The relatively dense mesh model of 468 vertices is well-suited for face-based AR effects. The proposed model demonstrates super-realtime inference speed on mobile GPUs (100-1000+ FPS, depending on the device and model variant) and a high prediction quality that is comparable to the variance in manual annotations of the same image.
32.MaskPlus: Improving Mask Generation for Instance Segmentation ⬇️
Instance segmentation is a promising yet challenging topic in computer vision. Recent approaches such as Mask R-CNN typically divide this problem into two parts -- a detection component and a mask generation branch, and mostly focus on the improvement of the detection part. In this paper, we present an approach that extends Mask R-CNN with five novel optimization techniques for improving the mask generation branch and reducing the conflicts between the mask branch and the detection component in training. These five techniques are independent to each other and can be flexibly utilized in building various instance segmentation architectures for increasing the overall accuracy. We demonstrate the effectiveness of our approach with tests on the COCO dataset.
33.Slow Feature Analysis for Human Action Recognition ⬇️
Slow Feature Analysis (SFA) extracts slowly varying features from a quickly varying input signal. It has been successfully applied to modeling the visual receptive fields of the cortical neurons. Sufficient experimental results in neuroscience suggest that the temporal slowness principle is a general learning principle in visual perception. In this paper, we introduce the SFA framework to the problem of human action recognition by incorporating the discriminative information with SFA learning and considering the spatial relationship of body parts. In particular, we consider four kinds of SFA learning strategies, including the original unsupervised SFA (U-SFA), the supervised SFA (S-SFA), the discriminative SFA (D-SFA), and the spatial discriminative SFA (SD-SFA), to extract slow feature functions from a large amount of training cuboids which are obtained by random sampling in motion boundaries. Afterward, to represent action sequences, the squared first order temporal derivatives are accumulated over all transformed cuboids into one feature vector, which is termed the Accumulated Squared Derivative (ASD) feature. The ASD feature encodes the statistical distribution of slow features in an action sequence. Finally, a linear support vector machine (SVM) is trained to classify actions represented by ASD features. We conduct extensive experiments, including two sets of control experiments, two sets of large scale experiments on the KTH and Weizmann databases, and two sets of experiments on the CASIA and UT-interaction databases, to demonstrate the effectiveness of SFA for human action recognition.
34.Natural Adversarial Examples ⬇️
We introduce natural adversarial examples -- real-world, unmodified, and naturally occurring examples that cause classifier accuracy to significantly degrade. We curate 7,500 natural adversarial examples and release them in an ImageNet classifier test set that we call ImageNet-A. This dataset serves as a new way to measure classifier robustness. Like l_p adversarial examples, ImageNet-A examples successfully transfer to unseen or black-box classifiers. For example, on ImageNet-A a DenseNet-121 obtains around 2% accuracy, an accuracy drop of approximately 90%. Recovering this accuracy is not simple because ImageNet-A examples exploit deep flaws in current classifiers including their over-reliance on color, texture, and background cues. We observe that popular training techniques for improving robustness have little effect, but we show that some architectural changes can enhance robustness to natural adversarial examples. Future research is required to enable robust generalization to this hard ImageNet test set.
35.Explaining Classifiers with Causal Concept Effect (CaCE) ⬇️
How can we understand classification decisions made by deep neural nets? We propose answering this question by using ideas from causal inference. We define the ``Causal Concept Effect'' (CaCE) as the causal effect that the presence or absence of a concept has on the prediction of a given deep neural net. We then use this measure as a mean to understand what drives the network's prediction and what does not. Yet many existing interpretability methods rely solely on correlations, resulting in potentially misleading explanations. We show how CaCE can avoid such mistakes. In high-risk domains such as medicine, knowing the root cause of the prediction is crucial. If we knew that the network's prediction was caused by arbitrary concepts such as the lighting conditions in an X-ray room instead of medically meaningful concept, this would prevent us from disastrous deployment of such models.
Estimating CaCE is difficult in situations where we cannot easily simulate the do-operator. As a simple solution, we propose learning a generative model, specifically a Variational AutoEncoder (VAE) on image pixels or image embeddings extracted from the classifier to measure VAE-CaCE. We show that VAE-CaCE is able to correctly estimate the true causal effect as compared to other baselines in controlled settings with synthetic and semi-natural high dimensional images.
36.Boosting Resolution and Recovering Texture of micro-CT Images with Deep Learning ⬇️
Digital Rock Imaging is constrained by detector hardware, and a trade-off between the image field of view (FOV) and the image resolution must be made. This can be compensated for with super resolution (SR) techniques that take a wide FOV, low resolution (LR) image, and super resolve a high resolution (HR), high FOV image. The Enhanced Deep Super Resolution Generative Adversarial Network (EDSRGAN) is trained on the Deep Learning Digital Rock Super Resolution Dataset, a diverse compilation 12000 of raw and processed uCT images. The network shows comparable performance of 50% to 70% reduction in relative error over bicubic interpolation. GAN performance in recovering texture shows superior visual similarity compared to SRCNN and other methods. Difference maps indicate that the SRCNN section of the SRGAN network recovers large scale edge (grain boundaries) features while the GAN network regenerates perceptually indistinguishable high frequency texture. Network performance is generalised with augmentation, showing high adaptability to noise and blur. HR images are fed into the network, generating HR-SR images to extrapolate network performance to sub-resolution features present in the HR images themselves. Results show that under-resolution features such as dissolved minerals and thin fractures are regenerated despite the network operating outside of trained specifications. Comparison with Scanning Electron Microscope images shows details are consistent with the underlying geometry of the sample. Recovery of textures benefits the characterisation of digital rocks with a high proportion of under-resolution micro-porous features, such as carbonate and coal samples. Images that are normally constrained by the mineralogy of the rock (coal), by fast transient imaging (waterflooding), or by the energy of the source (microporosity), can be super resolved accurately for further analysis downstream.
37.Anatomically-Informed Multiple Linear Assignment Problems for White Matter Bundle Segmentation ⬇️
Segmenting white matter bundles from human tractograms is a task of interest for several applications. Current methods for bundle segmentation consider either only prior knowledge about the relative anatomical position of a bundle, or only its geometrical properties. Our aim is to improve the results of segmentation by proposing a method that takes into account information about both the underlying anatomy and the geometry of bundles at the same time. To achieve this goal, we extend a state-of-the-art example-based method based on the Linear Assignment Problem (LAP) by including prior anatomical information within the optimization process. The proposed method shows a significant improvement with respect to the original method, in particular on small bundles.
38.CLCI-Net: Cross-Level fusion and Context Inference Networks for Lesion Segmentation of Chronic Stroke ⬇️
Segmenting stroke lesions from T1-weighted MR images is of great value for large-scale stroke rehabilitation neuroimaging analyses. Nevertheless, there are great challenges with this task, such as large range of stroke lesion scales and the tissue intensity similarity. The famous encoder-decoder convolutional neural network, which although has made great achievements in medical image segmentation areas, may fail to address these challenges due to the insufficient uses of multi-scale features and context information. To address these challenges, this paper proposes a Cross-Level fusion and Context Inference Network (CLCI-Net) for the chronic stroke lesion segmentation from T1-weighted MR images. Specifically, a Cross-Level feature Fusion (CLF) strategy was developed to make full use of different scale features across different levels; Extending Atrous Spatial Pyramid Pooling (ASPP) with CLF, we have enriched multi-scale features to handle the different lesion sizes; In addition, convolutional long short-term memory (ConvLSTM) is employed to infer context information and thus capture fine structures to address the intensity similarity issue. The proposed approach was evaluated on an open-source dataset, the Anatomical Tracings of Lesions After Stroke (ATLAS) with the results showing that our network outperforms five state-of-the-art methods. We make our code and models available at this https URL.
39.X-Net: Brain Stroke Lesion Segmentation Based on Depthwise Separable Convolution and Long-range Dependencies ⬇️
The morbidity of brain stroke increased rapidly in the past few years. To help specialists in lesion measurements and treatment planning, automatic segmentation methods are critically required for clinical practices. Recently, approaches based on deep learning and methods for contextual information extraction have served in many image segmentation tasks. However, their performances are limited due to the insufficient training of a large number of parameters, which sometimes fail in capturing long-range dependencies. To address these issues, we propose a depthwise separable convolution based X-Net that designs a nonlocal operation namely Feature Similarity Module (FSM) to capture long-range dependencies. The adopted depthwise convolution allows to reduce the network size, while the developed FSM provides a more effective, dense contextual information extraction and thus facilitates better segmentation. The effectiveness of X-Net was evaluated on an open dataset Anatomical Tracings of Lesions After Stroke (ATLAS) with superior performance achieved compared to other six state-of-the-art approaches. We make our code and models available at this https URL.
40.Single-bit-per-weight deep convolutional neural networks without batch-normalization layers for embedded systems ⬇️
Batch-normalization (BN) layers are thought to be an integrally important layer type in today's state-of-the-art deep convolutional neural networks for computer vision tasks such as classification and detection. However, BN layers introduce complexity and computational overheads that are highly undesirable for training and/or inference on low-power custom hardware implementations of real-time embedded vision systems such as UAVs, robots and Internet of Things (IoT) devices. They are also problematic when batch sizes need to be very small during training, and innovations such as residual connections introduced more recently than BN layers could potentially have lessened their impact. In this paper we aim to quantify the benefits BN layers offer in image classification networks, in comparison with alternative choices. In particular, we study networks that use shifted-ReLU layers instead of BN layers. We found, following experiments with wide residual networks applied to the ImageNet, CIFAR 10 and CIFAR 100 image classification datasets, that BN layers do not consistently offer a significant advantage. We found that the accuracy margin offered by BN layers depends on the data set, the network size, and the bit-depth of weights. We conclude that in situations where BN layers are undesirable due to speed, memory or complexity costs, that using shifted-ReLU layers instead should be considered; we found they can offer advantages in all these areas, and often do not impose a significant accuracy cost.
41.AirwayNet: A Voxel-Connectivity Aware Approach for Accurate Airway Segmentation Using Convolutional Neural Networks ⬇️
Airway segmentation on CT scans is critical for pulmonary disease diagnosis and endobronchial navigation. Manual extraction of airway requires strenuous efforts due to the complicated structure and various appearance of airway. For automatic airway extraction, convolutional neural networks (CNNs) based methods have recently become the state-of-the-art approach. However, there still remains a challenge for CNNs to perceive the tree-like pattern and comprehend the connectivity of airway. To address this, we propose a voxel-connectivity aware approach named AirwayNet for accurate airway segmentation. By connectivity modeling, conventional binary segmentation task is transformed into 26 tasks of connectivity prediction. Thus, our AirwayNet learns both airway structure and relationship between neighboring voxels. To take advantage of context knowledge, lung distance map and voxel coordinates are fed into AirwayNet as additional semantic information. Compared to existing approaches, AirwayNet achieved superior performance, demonstrating the effectiveness of the network's awareness of voxel connectivity.
42.Improved Reinforcement Learning through Imitation Learning Pretraining Towards Image-based Autonomous Driving ⬇️
We present a training pipeline for the autonomous driving task given the current camera image and vehicle speed as the input to produce the throttle, brake, and steering control output. The simulator Airsim's convenient weather and lighting API provides a sufficient diversity during training which can be very helpful to increase the trained policy's robustness. In order to not limit the possible policy's performance, we use a continuous and deterministic control policy setting. We utilize ResNet-34 as our actor and critic networks with some slight changes in the fully connected layers. Considering human's mastery of this task and the high-complexity nature of this task, we first use imitation learning to mimic the given human policy and leverage the trained policy and its weights to the reinforcement learning phase for which we use DDPG. This combination shows a considerable performance boost comparing to both pure imitation learning and pure DDPG for the autonomous driving task.
43.An Inter-Layer Weight Prediction and Quantization for Deep Neural Networks based on a Smoothly Varying Weight Hypothesis ⬇️
Network compression for deep neural networks has become an important part of deep learning research, because of increased demand for deep learning models in practical resource-constrained environments. In this paper, we observe that the weights in adjacent convolution layers share strong similarity in shapes and values, i.e., the weights tend to vary smoothly along the layers. We call this phenomenon \textit{Smoothly Varying Weight Hypothesis} (SVWH). Based on SVWH and an inter-frame prediction method in conventional video coding schemes, we propose a new \textit{Inter-Layer Weight Prediction} (ILWP) and quantization method which quantize the predicted residuals of the weights. Since the predicted weight residuals tend to follow Laplacian distributions with very low variance, the weight quantization can more effectively be applied, thus producing more zero weights and enhancing weight compression ratio. In addition, we propose a new loss for eliminating non-texture bits, which enabled us to more effectively store only texture bits. That is, the proposed loss regularizes the weights such that the collocated weights between the adjacent two layers have the same values. Our comprehensive experiments show that the proposed method achieved much higher weight compression rate at the same accuracy level compared with the previous quantization-based compression methods in deep neural networks.
44.Adversarial Sensor Attack on LiDAR-based Perception in Autonomous Driving ⬇️
In Autonomous Vehicles (AVs), one fundamental pillar is perception, which leverages sensors like cameras and LiDARs (Light Detection and Ranging) to understand the driving environment. Due to its direct impact on road safety, multiple prior efforts have been made to study its the security of perception systems. In contrast to prior work that concentrates on camera-based perception, in this work we perform the first security study of LiDAR-based perception in AV settings, which is highly important but unexplored. We consider LiDAR spoofing attacks as the threat model and set the attack goal as spoofing obstacles close to the front of a victim AV. We find that blindly applying LiDAR spoofing is insufficient to achieve this goal due to the machine learning-based object detection process. Thus, we then explore the possibility of strategically controlling the spoofed attack to fool the machine learning model. We formulate this task as an optimization problem and design modeling methods for the input perturbation function and the objective function. We also identify the inherent limitations of directly solving the problem using optimization and design an algorithm that combines optimization and global sampling, which improves the attack success rates to around 75%. As a case study to understand the attack impact at the AV driving decision level, we construct and evaluate two attack scenarios that may damage road safety and mobility. We also discuss defense directions at the AV system, sensor, and machine learning model levels.
45.Deep learning-based color holographic microscopy ⬇️
We report a framework based on a generative adversarial network (GAN) that performs high-fidelity color image reconstruction using a single hologram of a sample that is illuminated simultaneously by light at three different wavelengths. The trained network learns to eliminate missing-phase-related artifacts, and generates an accurate color transformation for the reconstructed image. Our framework is experimentally demonstrated using lung and prostate tissue sections that are labeled with different histological stains. This framework is envisaged to be applicable to point-of-care histopathology, and presents a significant improvement in the throughput of coherent microscopy systems given that only a single hologram of the specimen is required for accurate color imaging.