Generating new visual tasks requires additional datasets, and it costs a considerable effort. We propose a new method of domain adaptation for merging multiple models with less effort than creating an additional dataset. This method merges pre-trained models in different domains using the glue layers and the generative model, which feeds latent features to train the glue layers without an additional dataset. We also propose the generative model created by knowledge distillation from pre-trained models. It also allows reusing the dataset to create latent features for training the glue layers. We apply this method to object detection in a low-light situation. The “YOLO in the Dark” contains two models, “Learning to See in the Dark” and YOLO. We report the new method and the result of domain adaptation that detect objects from raw short-exposure low-light images. The “YOLO in the Dark” costs fewer computing resources compared to the naive approach.
论文检索
输入标题、作者或关键词，从 1,358 篇学术成果中精准定位
Zero-shot domain adaptation (ZSDA) is a category of domain adaptation problems where neither data sample nor label is available for parameter learning in the target domain. With the hypothesis that the shift between a given pair of domains is shared across tasks, we propose a new method for ZSDA by transferring domain shift from an irrelevant task (IrT) to the task of interest (ToI). Specifically, we first identify an IrT, where dual-domain samples are available, and capture the domain shift with a coupled generative adversarial networks (CoGAN) in this task. Then, we train a CoGAN for the ToI and restrict it to carry the same domain shift as the CoGAN for IrT does. In addition, we introduce a pair of co-training classifiers to regularize the training procedure of CoGAN in the ToI. The proposed method not only derives machine learning models for the non-available target-domain data, but also synthesizes the data themselves. We evaluate the proposed method on benchmark datasets and achieve the state-of-the-art performances.
Tehchniques for manipulating images are advancing rapidly; while these are helpful for many useful tasks, they also pose a threat to society with their ability to create believable misinformation. We present a novel, Spatial Pyramid Attention Network (SPAN) for detection and localization of multiple types of image manipulations. The proposed architecture efficiently and effecively models the relationship between image patches at multiple scales by constructing a pyramid of local self-attention blocks. The design includes a novel position projection to encode the spatial positions of the patches. SPAN is trained on a synthetic dataset but can also be fine tuned for specific datasets; The proposed method shows significant gains in performance on standard datasets over previous state-of-art methods.
We introduce a differential visual similarity metric to train deep neural networks for 3D reconstruction, aimed at improving reconstruction quality. The metric compares two 3D shapes by measuring distances between multi-view images differentiably rendered from the shapes. Importantly, the image-space distance is also differentiable and measures visual similarity, rather than pixel-wise distortion. Specifically, the similarity is defined by mean-squared errors over HardNet features computed from probabilistic keypoint maps of the compared images. Our differential visual shape similarity metric can be easily plugged into various 3D reconstruction networks, replacing their distortion-based losses, such as Chamfer or Earth Mover distances, so as to optimize the network weights to produce reconstructions with better structural fidelity and visual quality. We demonstrate this both objectively, using well-known shape metrics for retrieval and classification tasks that are independent from our new metric, and subjectively through a perceptual study.
Data augmentation has been widely adopted for object detection in 3D point clouds. However, all previous related efforts have focused on manually designing specific data augmentation methods for individual architectures. In this work, we present the first attempt to automate the design of data augmentation policies for 3D object detection. We introduce the Progressive Population Based Augmentation (PPBA) algorithm, which learns to optimize augmentation strategies by narrowing down the search space and adopting the best parameters discovered in previous iterations. On the KITTI 3D detection test set, PPBA improves the StarNet detector by substantial margins on the moderate difficulty category of cars, pedestrians, and cyclists, outperforming all current state-of-the-art single-stage detection models. Additional experiments on the Waymo Open Dataset indicate that PPBA continues to effectively improve the StarNet and PointPillars detectors on a 20x larger dataset compared to KITTI. The magnitude of the improvements may be comparable to advances in 3D perception architectures and the gains come without an incurred cost at inference time. In subsequent experiments, we find that PPBA may be up to 10x more data efficient than baseline 3D detection models without augmentation, highlighting that 3D detection models may achieve competitive accuracy with far fewer labeled examples.
Recently, there is an increasing interest in highlight detection research where the goal is to create a short duration video from a longer video by extracting its interesting moments. However, most existing methods ignore the fact that the definition of video highlight is highly subjective. Different users may have different preferences of highlight for the same input video. In this paper, we propose a simple yet effective framework that learns to adapt highlight detection to a user by exploiting the user's history in the form of highlights that the user has previously created. Our framework consists of two sub-networks: a fully temporal convolutional highlight detection network $H$ that predicts highlight for an input video and a history encoder network $M$ for user history. We introduce a newly designed temporal-adaptive instance normalization (T-AIN) layer to $H$ where the two sub-networks interact with each other. T-AIN has affine parameters that are predicted from $M$ based on the user history and is responsible for the user-adaptive signal to $H$. Extensive experiments on a large-scale dataset show that our framework can make more accurate and user-specific highlight predictions.
Dropout has been widely adopted to regularize graph convolutional networks (GCNs) by randomly zeroing entries of the node feature vectors and obtains promising performance on various tasks. However, the information of individually zeroed entries could still present in other correlated entries by propagating (1) spatially between entries of different node feature vectors and (2) depth-wisely between different entries of each node feature vector, which essentially weakens the effectiveness of dropout. This is mainly because in a GCN, neighboring node feature vectors after linear transformations are aggregated to produce new node feature vectors in the subsequent layer. To effectively regularize GCNs, we devise DropCluster which first randomly zeros some seed entries and then zeros entries that are spatially or depth-wisely correlated to those seed entries. In this way, the information of the seed entries is thoroughly removed and cannot flow to subsequent layers via the correlated entries. We validate the effectiveness of the proposed DropCluster by comprehensively comparing it with dropout and its representative variants, such as SpatialDropout, Gaussian dropout and DropEdge, on skeleton-based action recognition.
We present a new object representation, called Dense Rep-Points, which utilize a large number of points to describe the multi-grainedobject representation of both box level and pixel level. Techniques are pro-posed to efficiently process these dense points, which maintains nearconstant complexity with increasing point number. The dense RepPointsis proved to represent and learn object segment well, by a novel dis-tance transform sampling method combined with a set-to-set supervision.The novel distance transform sampling method combines the strengthof contour and grid representation, which significantly outperforms thecounter-parts using contour or grid representations. On COCO, it achieves39.6 mask AP and 48.3 bbox AP.
Learning representations that clearly distinguish between normal and abnormal data is key to the success of anomaly detection. Most of existing anomaly detection algorithms use activation representations from forward propagation while not exploiting gradients from backpropagation to characterize data. Gradients capture model updates required to represent data. Anomalies require more drastic model updates to fully represent them compared to normal data. Hence, we propose the utilization of backpropagated gradients as representations to characterize model behavior on anomalies and, consequently, detect such anomalies. We show that the proposed method using gradient-based representations achieves state-of-the-art anomaly detection performance in benchmark image recognition datasets. Also, we highlight the computational efficiency and the simplicity of the proposed method in comparison with other state-of-the-art methods relying on adversarial networks or autoregressive models, which require at least 27 times more model parameters than the proposed method.
Today's computer-aided diagnosis (CAD) model is still far from the clinical practice of glaucoma detection, mainly due to the training bias originating from 1) the normal-abnormal class imbalance and 2) the rare but significant hard samples in fundus images. However, debiasing in CAD is not trivial because existing methods cannot cure the two types of bias to categorize fundus images. In this paper, we propose a novel curriculum learning paradigm (EGDCL) to train an unbiased glaucoma diagnosis model with the adaptive dual-curriculum. Innovatively, the dual-curriculum is designed with the guidance of evidence maps to build a training criterion, which gradually cures the bias in training data. In particular, the dual-curriculum emphasizes unbiased training contributions of data from easy to hard, normal to abnormal, and the dual-curriculum is optimized jointly with model parameters to obtain the optimal solution. In comparison to baselines, EGDCL significantly improves the convergence speed of the training process and obtains the top performance in the test procedure. Experimental results on challenging glaucoma datasets show that our EGDCL delivers unbiased diagnosis (0.9721 of Sensitivity, 0.9707 of Specificity, 0.993 of AUC, 0.966 of F2-score) and outperform the other methods. It endows our EGDCL a great advantage to handle the unbiased CAD in clinical application.
Knowledge distillation refers to the process of training a student network to achieve better accuracy by learning from a pre-trained teacher network. Most of the existing knowledge distillation methods direct the student to follow the teacher by matching the teacher's output, feature maps or their distribution. In this work, we propose a novel way to achieve this goal: by distilling the knowledge through a quantized visual words space. According to our method, the teacher's feature maps are first quantized to represent the main visual concepts (i.e., visual words) encompassed in these maps and then the student is asked to predict those visual word representations. Despite its simplicity, we show that our approach is able to yield results that improve the state of the art on knowledge distillation for model compression and transfer learning scenarios. To that end, we provide an extensive evaluation across several network architectures and most commonly used benchmark datasets.
In this paper, we propose the Deep Structured self-Driving Network (DSDNet), which performs object detection, motion prediction, and motion planning with a single neural network. Towards this goal, we develop a deep structured energy based model which considers the interactions between actors and produces socially consistent multimodal future predictions. Furthermore, DSDNet explicitly exploits the predicted future distributions of actors to plan a safe maneuver by using a structured planning cost. Our sample-based formulation allows us to overcome the difficulty in probabilistic inference of continuous random variables. Experiments on a number of large-scale self driving datasets demonstrate that our model significantly outperforms the state-of-the-art.
Monocular depth estimation algorithms successfully predict the relative depth order of objects in a scene. However, because of the fundamental scale ambiguity associated with monocular images, these algorithms fail at correctly predicting true metric depth. In this work, we demonstrate how a depth histogram of the scene, which can be readily captured using a single-pixel time-resolved detector, can be fused with the output of existing monocular depth estimation algorithms to resolve the depth ambiguity problem. We validate this novel sensor fusion technique experimentally and in extensive simulation. We show that it significantly improves the performance of several state-of-the-art monocular depth estimation algorithms.
Imaging in low light is difficult because the number of photons arriving at the sensor is low. Imaging dynamic scenes in low-light environments is even more difficult because as the scene moves, pixels in adjacent frames need to be aligned before they can be denoised. Conventional CMOS image sensors (CIS) are at a particular disadvantage in dynamic low-light settings because the exposure cannot be too short lest the read noise overwhelms the signal. We propose a solution using Quanta Image Sensors (QIS) and present a new image reconstruction algorithm. QIS are single-photon image sensors with photon counting capabilities. Studies over the past decade have confirmed the effectiveness of QIS for low-light imaging but reconstruction algorithms for dynamic scenes in low light remain an open problem. We fill the gap by proposing a student-teacher training protocol that transfers knowledge from a motion teacher and a denoising teacher to a student network. We show that dynamic scenes can be reconstructed from a burst of frames at a photon level of 1 photon per pixel per frame. Experimental results confirm the advantages of the proposed method compared to existing methods.
Multi-view stereo (MVS) is the golden mean between the accuracy of active depth sensing and the practicality of monocular depth estimation. Cost volume based approaches employing 3D convolutional neural networks (CNNs) have considerably improved the accuracy of MVS systems. However, this accuracy comes at a high computational cost which impedes practical adoption. Distinct from cost volume approaches, we propose an efficient depth estimation approach by first (a) detecting and evaluating descriptors for interest points, then (b) learning to match and triangulate a small set of interest points, and finally densifying this sparse set of 3D points using CNNs. An end-to-end network efficiently performs all three steps within a deep learning framework and trained with intermediate 2D image and 3D geometric supervision, along with depth supervision. Crucially, our first step complements pose estimation using interest point detection and descriptor learning. We demonstrate state-of-the-art results on depth estimation with lower compute for different scene lengths. Furthermore, our method generalizes to newer environments and the descriptors output by our network compare favorably to strong baselines.
Place is an important element in visual understanding. Given a photo of a building, people can often tell its functionality, e.g. a restaurant or a shop, its cultural style, e.g. Asian or European, as well as its economic type, e.g. industry oriented or tourism oriented. While place recognition has been widely studied in previous work, there remains a long way towards comprehensive place understanding, which is far beyond categorizing a place with an image and requires information of multiple aspects. In this work, we contribute Placepedia1, a large-scale place dataset with more than 35M photos from 240K unique places. Besides the photos, each place also comes with massive multi-faceted information, e.g. GDP, population, etc., and labels at multiple levels, including function, city, country, etc.. This dataset, with its large amount of data and rich annotations, allows various studies to be conducted. Particularly, in our studies, we develop 1) PlaceNet, a uni ed framework for multi-level place recognition, and 2) a method for city embedding, which can produce a vector representation for a city that captures both visual and multi-faceted side information. Such studies not only reveal key challenges in place understanding, but also establish connections between visual observations and underlying socioeconomic/cultural implications.
Accurate 3D object detection in LiDAR based point clouds suffers from the challenges of data sparsity and irregularities. Existing methods strive to organize the points regularly, e.g. voxelize, pass them through a designed 2D/3D neural network, and then define object-level anchors that predict offsets of 3D bounding boxes using collective evidences from all the points on the objects of interest. Contrary to the state-of-the-art anchor-based methods, based on the very nature of data sparsity, we observe that even points on an individual object part are informative about semantic information of the object. We thus argue in this paper for an approach opposite to existing methods using object-level anchors. Inspired by compositional models, which represent an object as parts and their spatial relations, we propose to represent an object as composition of its interior non-empty voxels, termed hotspots, and the spatial relations of hotspots. This gives rise to the representation of Object as Hotspots (OHS). Based on OHS, we further propose an anchor-free detection head with a novel ground truth assignment strategy that deals with inter-object point-sparsity imbalance to prevent the network from biasing towards objects with more points. Experimental results show that our proposed method works remarkably well on objects with a small number of points. Notably, our approach ranked 1st on KITTI 3D Detection Benchmark for cyclist and pedestrian detection, and achieved state-of-the-art performance on NuScenes 3D Detection Benchmark.
We introduce the task of Image-Set Visual Question Answering (ISVQA), which generalizes the commonly studied single-image VQA problem to multi-image settings. Taking a natural language question and a set of images as input, it aims to answer the question based on the content of the images. The questions can be about objects and relationships in one or more images or about the entire scene depicted by the image set. To enable research in this new topic, we introduce two ISVQA datasets - indoor and outdoor scenes. They simulate the real-world scenarios of indoor image collections and multiple car-mounted cameras, respectively. The indoor-scene dataset contains 91,479 human-annotated questions for 48,138 image sets, and the outdoor-scene dataset has 49,617 questions for 12,746 image sets. We analyze the properties of the two datasets, including question-and-answer distributions, types of questions, biases in dataset, and question-image dependencies. We also build new baseline models to investigate new research challenges in ISVQA.
Physical adversarial attacks threaten to fool object detection systems, but reproducible research on the real-world effectiveness of physical patches and how to defend against them requires a publicly available benchmark dataset. We present APRICOT, a collection of over 1,000 annotated photographs of printed adversarial patches in public locations. The patches target several object categories for three COCO-trained detection models, and the photos represent natural variation in position, distance, lighting conditions, and viewing angle. Our analysis suggests that maintaining adversarial robustness in uncontrolled settings is highly challenging, but it is still possible to produce targeted detections under white-box and sometimes black-box settings. We establish baselines for defending against adversarial patches through several methods, including a detector supervised with synthetic data and unsupervised methods such as kernel density estimation, Bayesian uncertainty, and reconstruction error. Our results suggest that adversarial patches can be effectively flagged, both in a high-knowledge, attack-specific scenario, and in an unsupervised setting where patches are detected as anomalies in natural images. This dataset and the described experiments provide a benchmark for future research on the effectiveness of and defenses against physical adversarial objects in the wild.
Autoregressive models recently achieved comparable results versus state-of-the-art Generative Adversarial Networks (GANs) with the help of Vector Quantized Variational AutoEncoders (VQ-VAE). However, autoregressive models have several limitations such as exposure bias and their training objective does not guarantee visual fidelity. To address these limitations, we propose to use Reinforced Adversarial Learning (RAL) based on policy gradient optimization for autoregressive models. By applying RAL, we enable a similar process for training and testing to address the exposure bias issue. In addition, visual fidelity has been further optimized with adversarial loss inspired by their strong counterparts: GANs. Due to the slow sampling speed of autoregressive models, we propose to use partial generation for faster training. RAL also empowers the collaboration between different modules of the VQ-VAE framework. To our best knowledge, the proposed method is first to enable adversarial learning in autoregressive models for image generation. Experiments on synthetic and real-world datasets show improvements over the MLE trained models. The proposed method improves both negative log-likelihood (NLL) and Fréchet Inception Distance (FID), which indicates improvements in terms of visual quality and diversity. The proposed method achieves state-of-the-art results on Celeba for 64x64 image resolution, showing promise for large scale image generation.