Viewpoint Estimation Research Articles

SHOWMe: Robust object-agnostic hand-object 3D reconstruction from RGB video

In this paper, we tackle the problem of detailed hand-object 3D reconstruction from monocular video with unknown objects, for applications where the required accuracy and level of detail is important, e.g. object hand-over in human–robot collaboration, or manipulation and contact point analysis. While the recent literature on this topic is promising, the accuracy and generalization abilities of existing methods are still lacking. This is due to several limitations, such as the assumption of known object class or model for a small number of instances, or over-reliance on off-the-shelf keypoint and structure-from-motion methods for object-relative viewpoint estimation, prone to complete failure with previously unobserved, poorly textured objects or hand-object occlusions. To address previous method shortcomings, we present a 2-stage pipeline superseding state-of-the-art (SotA) performance on several metrics. First, we robustly retrieve viewpoints relying on a learned pairwise camera pose estimator trainable with a low data regime, followed by a globalized Shonan pose averaging. Second, we simultaneously estimate detailed 3D hand-object shapes and refine camera poses using a differential renderer-based optimizer. To better assess the out-of-distribution abilities of existing methods, and to showcase our methodological contributions, we introduce the new SHOWMe benchmark dataset with 96 sequences annotated with poses, millimetric textured 3D shape scans, and parametric hand models, introducing new object and hand diversity. Remarkably, we show that our method is able to reconstruct 100% of these sequences as opposed to SotA Structure-from-Motion (SfM) or hand-keypoint-based pipelines, and obtains reconstructions of equivalent or better precision when existing methods do succeed in providing a result. We hope these contributions lead to further research under harder input assumptions. The dataset can be downloaded at https://download.europe.naverlabs.com/showme.

Grounding human-object interaction to affordance behavior in multimodal datasets.

While affordance detection and Human-Object interaction (HOI) detection tasks are related, the theoretical foundation of affordances makes it clear that the two are distinct. In particular, researchers in affordances make distinctions between J. J. Gibson's traditional definition of an affordance, "the action possibilities" of the object within the environment, and the definition of a telic affordance, or one defined by conventionalized purpose or use. We augment the HICO-DET dataset with annotations for Gibsonian and telic affordances and a subset of the dataset with annotations for the orientation of the humans and objects involved. We then train an adapted Human-Object Interaction (HOI) model and evaluate a pre-trained viewpoint estimation system on this augmented dataset. Our model, AffordanceUPT, is based on a two-stage adaptation of the Unary-Pairwise Transformer (UPT), which we modularize to make affordance detection independent of object detection. Our approach exhibits generalization to new objects and actions, can effectively make the Gibsonian/telic distinction, and shows that this distinction is correlated with features in the data that are not captured by the HOI annotations of the HICO-DET dataset.

Lightweight and Effective Convolutional Neural Networks for Vehicle Viewpoint Estimation From Monocular Images

Vehicle viewpoint estimation from monocular images is a crucial component for autonomous driving vehicles and for fleet management applications. In this paper, we make several contributions to advance the state-of-the-art on this problem. We show the effectiveness of applying a smoothing filter to the output neurons of a Convolutional Neural Network (CNN) when estimating vehicle viewpoint. We point out the overlooked fact that, under the same viewpoint, the appearance of a vehicle is strongly influenced by its position in the image plane, which renders viewpoint estimation from appearance an ill-posed problem. We show how, by inserting in the model a CoordConv layer to provide the coordinates of the vehicle, we are able to solve such ambiguity and greatly increase performance. Finally, we introduce a new data augmentation technique that improves viewpoint estimation on vehicles that are closer to the camera or partially occluded. All these improvements let a lightweight CNN reach optimal results while keeping inference time low. An extensive evaluation on a viewpoint estimation benchmark (Pascal3D+) and on actual vehicle camera data (nuScenes) shows that our method significantly outperforms the state-of-the-art in vehicle viewpoint estimation, both in terms of accuracy and memory footprint.

When and how convolutional neural networks generalize to out-of-distribution category–viewpoint combinations

Object recognition and viewpoint estimation lie at the heart of visual understanding. Recent studies have suggested that convolutional neural networks (CNNs) fail to generalize to out-of-distribution (OOD) category–viewpoint combinations, that is, combinations not seen during training. Here we investigate when and how such OOD generalization may be possible by evaluating CNNs trained to classify both object category and three-dimensional viewpoint on OOD combinations, and identifying the neural mechanisms that facilitate such OOD generalization. We show that increasing the number of in-distribution combinations (data diversity) substantially improves generalization to OOD combinations, even with the same amount of training data. We compare learning category and viewpoint in separate and shared network architectures, and observe starkly different trends on in-distribution and OOD combinations, that is, while shared networks are helpful in distribution, separate networks significantly outperform shared ones at OOD combinations. Finally, we demonstrate that such OOD generalization is facilitated by the neural mechanism of specialization, that is, the emergence of two types of neuron—neurons selective to category and invariant to viewpoint, and vice versa.

Few-Shot Object Detection and Viewpoint Estimation for Objects in the Wild.

Detecting objects and estimating their viewpoints in images are key tasks of 3D scene understanding. Recent approaches have achieved excellent results on very large benchmarks for object detection and viewpoint estimation. However, performances are still lagging behind for novel object categories with few samples. In this paper, we tackle the problems of few-shot object detection and few-shot viewpoint estimation. We demonstrate on both tasks the benefits of guiding the network prediction with class-representative features extracted from data in different modalities: image patches for object detection, and aligned 3D models for viewpoint estimation. Despite its simplicity, our method outperforms state-of-the-art methods by a large margin on a range of datasets, including PASCAL and COCO for few-shot object detection, and Pascal3D+ and ObjectNet3D for few-shot viewpoint estimation. Furthermore, when the 3D model is not available, we introduce a simple category-agnostic viewpoint estimation method by exploiting geometrical similarities and consistent pose labeling across different classes. While it moderately reduces performance, this approach still obtains better results than previous methods in this setting. Last, for the first time, we tackle the combination of both few-shot tasks, on three challenging benchmarks for viewpoint estimation in the wild, ObjectNet3D, Pascal3D+ and Pix3D, showing very promising results.

See360: Novel Panoramic View Interpolation.

We present See360, which is a versatile and efficient framework for 360° panoramic view interpolation using latent space viewpoint estimation. Most of the existing view rendering approaches only focus on indoor or synthetic 3D environments and render new views of small objects. In contrast, we suggest to tackle camera-centered view synthesis as a 2D affine transformation without using point clouds or depth maps, which enables an effective 360° panoramic scene exploration. Given a pair of reference images, the See360 model learns to render novel views by a proposed novel Multi-Scale Affine Transformer (MSAT), enabling the coarse-to-fine feature rendering. We also propose a Conditional Latent space AutoEncoder (C-LAE) to achieve view interpolation at any arbitrary angle. To show the versatility of our method, we introduce four training datasets, namely UrbanCity360, Archinterior360, HungHom360 and Lab360, which are collected from indoor and outdoor environments for both real and synthetic rendering. Experimental results show that the proposed method is generic enough to achieve real-time rendering of arbitrary views for all four datasets. In addition, our See360 model can be applied to view synthesis in the wild: with only a short extra training time (approximately 10 mins), and is able to render unknown real-world scenes. The superior performance of See360 opens up a promising direction for camera-centered view rendering and 360° panoramic view interpolation.

Adversarial Learning for Depth and Viewpoint Estimation From a Single Image

Estimating a depth map and, at the same time, predicting the 3D pose of an object from a single 2D color image is a very challenging task. Depth estimation is typically performed through stereo vision by following several time-consuming stages, such as epipolar geometry, rectification and matching. Alternatively, when stereo vision is not useful or applicable, depth relations can be inferred from a single image as studied in this paper. More precisely, deep learning is applied in order to solve the problem of estimating a depth map from a single image. Then, that map is used for predicting the 3D pose of the main object depicted in the image. The proposed model consists of two successive neural networks. The first network is based on a Generative Adversarial Neural network (GAN). It estimates a dense depth map from the given color image. A Convolutional Neural Network (CNN) is then used to predict the 3D pose from the generated depth map through regression. The main difficulty to jointly estimate depth maps and 3D poses using deep networks is the lack of training data with both depth and viewpoint annotations. This contribution assumes a cross-domain training procedure with 3D CAD models corresponding to objects appearing in real images in order to render depth images from different viewpoints. These rendered images are then used to guide the GAN network to learn the mapping from the image domain to the depth domain. By exploiting the dataset as a source of training data, the proposed model outperforms state-of-the-art models on the PASCAL 3D+ dataset. The code of the proposed model is publicly available at https://github.com/SaddamAbdulrhman/Depth-and-Viewpoint-Estimation/tree/master.

EasyMesh: An efficient method to reconstruct 3D mesh from a single image

How to reconstruct the complete 3D mesh model from a single natural image is now still considered as a challenging problem. Most existing methods describe 3D shapes in the form of voxel or point cloud, and it is not always trivial to convert them into meshes with high quality. In this paper, we present a novel method to effectively address this problem by using a specially-designed GAN model to map a given natural image to a geometry image, from which the corresponding 3D mesh can be reconstructed. Specifically, we disentangle the tasks of viewpoint estimation and 3D reconstruction, ensuring that the reconstruction network focuses on generating vivid 3D meshes with accurate viewpoint information. We also add a differentiable module to create silhouettes from various viewpoints for the synthesized geometry image, aiming to improve the consistency between the generated 3D model and its input 2D image. Furthermore, we design a compact but effective discriminator for geometry images to guarantee a plausible overall contour of the generated object. Experiments conducted on a publicly available database demonstrate that the proposed method can generate 3D meshes with high fidelity and outperforms other state-of-the-art approaches in both qualitative and quantitative results. Our code is publicly available at https://github.com/tasx0823/EasyMesh.

A Viewpoint Aware Multi-Task Learning Framework for Fine-Grained Vehicle Recognition

The goal of fine-grained vehicle recognition is to identify the exact subtype of the vehicle from a given image. It plays an important role in the intelligent traffic surveillance system. Although fine-grained vehicle recognition has attracted more and more research interest, it remains as an open problem for vehicle images taken from arbitrary viewpoints. In this study, we present a one-stage deep multi-task learning framework for fine-grained vehicle recognition in traffic surveillance, which performs the fine-grained vehicle recognition and viewpoint estimation simultaneously. In the proposed framework, the fine-grained vehicle recognition is the main task which classifies images into different types, and the viewpoint estimation task is the auxiliary task to learn helpful viewpoint-aware features for improving the main task. We evaluate our method on two common used large-scale fine-grained vehicle recognition datasets, including BoxCars116k and CompCars. The experimental results testify that the proposed multi-task framework can improve the accuracy by incorporating the viewpoint information of the vehicles. In comparison to the state-of-the-art, our approach goes beyond, only requiring 3D bounding box for the training phase, which is important for future inferences using the trained model.

Weakly-Supervised Learning of Category-Specific 3D Object Shapes.

Category-specific 3D object shape models have greatly boosted the recent advances in object detection, recognition and segmentation. However, even the most advanced approach for learning 3D object shapes still requires heavy manual annotations on large-scale 2D images. Such annotations include object categories, object keypoints, and figure-ground segmentation for the instances in each image. In particular, annotating figure-ground segmentation is unbearably labor-intensive and time-consuming. To address this problem, this paper devotes to learn category-specific 3D shape models under weak supervision, where only object categories and keypoints are required to be manually annotated on the training 2D images. By exploring the underlying relationship between two tasks: object segmentation and category-specific 3D shape reconstruction, we propose a novel weakly-supervised learning framework to jointly address these two tasks and combine them to boost the final performance of the learned 3D shape models. Moreover, learning without using figure-ground segmentation leads to ambiguous solutions. To this end, we develop the confidence weighting schemes in the viewpoint estimation and 3D shape learning procedure. These schemes effectively reduce the confusion caused by the noisy data and thus increase the chances for recovering more reliable 3D object shapes. Comprehensive experiments on the challenging PASCAL VOC benchmark show that our framework achieves comparable performance with the state-of-the-art methods that use expensive manual segmentation-level annotations. In addition, our experiments also demonstrate that our 3D shape models improve object segmentation performance.

Boosting Multi-Vehicle Tracking with a Joint Object Detection and Viewpoint Estimation Sensor.

In this work, we address the problem of multi-vehicle detection and tracking for traffic monitoring applications. We preset a novel intelligent visual sensor for tracking-by-detection with simultaneous pose estimation. Essentially, we adapt an Extended Kalman Filter (EKF) to work not only with the detections of the vehicles but also with their estimated coarse viewpoints, directly obtained with the vision sensor. We show that enhancing the tracking with observations of the vehicle pose, results in a better estimation of the vehicles trajectories. For the simultaneous object detection and viewpoint estimation task, we present and evaluate two independent solutions. One is based on a fast GPU implementation of a Histogram of Oriented Gradients (HOG) detector with Support Vector Machines (SVMs). For the second, we adequately modify and train the Faster R-CNN deep learning model, in order to recover from it not only the object localization but also an estimation of its pose. Finally, we publicly release a challenging dataset, the GRAM Road Traffic Monitoring (GRAM-RTM), which has been especially designed for evaluating multi-vehicle tracking approaches within the context of traffic monitoring applications. It comprises more than 700 unique vehicles annotated across more than 40.300 frames of three videos. We expect the GRAM-RTM becomes a benchmark in vehicle detection and tracking, providing the computer vision and intelligent transportation systems communities with a standard set of images, annotations and evaluation procedures for multi-vehicle tracking. We present a thorough experimental evaluation of our approaches with the GRAM-RTM, which will be useful for establishing further comparisons. The results obtained confirm that the simultaneous integration of vehicle localizations and pose estimations as observations in an EKF, improves the tracking results.

3D Wireframe Modeling and Viewpoint Estimation for Multi-Class Objects Combining Deep Neural Network and Deformable Model Matching

The accuracy of 3D viewpoint and shape estimation from 2D images has been greatly improved by machine learning, especially deep learning technology such as the convolution neural network (CNN). However, current methods are always valid only for one specific category and have exhibited poor performance when generalized to other categories, which means that multiple detectors or networks are needed for multi-class object image cases. In this paper, we propose a method with strong generalization ability, which incorporates only one CNN with deformable model matching processing for the 3D viewpoint and the shape estimation of multi-class object image cases. The CNN is utilized to detect keypoints of the potential object from the image, while a deformable model matching stage is designed to conduct 3D wireframe modeling and viewpoint estimation simultaneously with the support of the detected keypoints. Besides, parameter estimation by deformable model matching processing has robust fault-tolerance to the keypoint detection results containing mistaken keypoints. The proposed method is evaluated on Pascal3D+ dataset. Experiments show that the proposed method performs well in both parameter estimation accuracy and the multi-class objects generalization. This research is a useful exploration to extend the generalization of deep learning in specific tasks.

CNNs Based Viewpoint Estimation for Volume Visualization

Viewpoint estimation from 2D rendered images is helpful in understanding how users select viewpoints for volume visualization and guiding users to select better viewpoints based on previous visualizations. In this article, we propose a viewpoint estimation method based on Convolutional Neural Networks (CNNs) for volume visualization. We first design an overfit-resistant image rendering pipeline to generate the training images with accurate viewpoint annotations, and then train a category-specific viewpoint classification network to estimate the viewpoint for the given rendered image. Our method can achieve good performance on images rendered with different transfer functions and rendering parameters in several categories. We apply our model to recover the viewpoints of the rendered images in publications, and show how experts look at volumes. We also introduce a CNN feature-based image similarity measure for similarity voting based viewpoint selection, which can suggest semantically meaningful optimal viewpoints for different volumes and transfer functions.

Viewpoint Estimation using Triplet Loss with A Novel Viewpoint-based Input Selection Strategy

Viewpoint estimation is a fundamental procedure in vision-based robot tasks. A good viewpoint of the camera relative to objects can help the visual system perform better both in observation and manipulation. Recently, CNN-based algorithms, which can effectively extract discriminative features from images in challenging conditions, are utilized to handle the viewpoint estimation problem. However, most existing algorithms focus on how to leverage the extracted deep features while neglecting the spatial relationship among images that captured from various viewpoints. In this paper, we present a deep metric learning method for solving the viewpoint estimation problem. A triplet loss with a novel viewpoint-based input selection strategy is introduced, which could learn more powerful features after incorporating the spatial relationship between viewpoints. Combined with the traditional classification loss, the presented loss can further enhance the discriminative power of features. To evaluate the performance of our method, a dataset containing a large number of images generated from five different texture-less workpieces is built and the experiment results show the effectiveness of the proposed method.

Human Detection and Motion Analysis from a Quadrotor UAV

This work focuses on detecting humans and estimating their pose and trajectory from an umnanned aerial vehicle (UAV). In our framework, a human detection model is trained using a Region-based Convolutional Neural Network (R-CNN). Each video frame is corrected for perspective using projective transformation. Using Histogram Oriented Gradients (HOG) of the silhouettes as features, the detected human figures are then classified for their pose. A dynamic classifier is developed to estimate forward walking and a turning gait sequence. The estimated poses are used to estimate the shape of the trajectory traversed by the human subject. An average precision of 98% has been achieved for the detector. Experiments conducted on aerial videos confirm our solution can achieve accurate pose and trajectory estimation for different kinds of perspective-distorted videos. For example, for a video recorded at 40m above ground, the perspective correction improves accuracy by 37.1% and 17.8% in pose and viewpoint estimation respectively.

Viewpoint refinement and estimation with adapted synthetic data

Estimating the viewpoint of objects in images is an important task for scene understanding. The viewpoint estimation accuracy, however, depends highly on the amount of training data and the quality of the annotation. While humans excel at labelling images with coarse viewpoint annotations like front, back, left or right, the process becomes tedious and the quality of the annotations decreases when finer viewpoint discretisations are required. To solve this problem, we propose a refinement of coarse viewpoint annotations, which are provided by humans, with synthetic data automatically generated from 3D models. To compensate between the difference between synthetic and real images, we introduce a domain adaptation approach that aligns the domain of the synthesized images with the domain of the real images. Experiments show that the proposed approach significantly improves viewpoint estimation on several state-of-the-art datasets.

Fast Joint Object Detection and Viewpoint Estimation for Traffic Scene Understanding

Environment perception is a critical enabler for automated driving systems since it allows a comprehensive understanding of traffic situations, which is a requirement to ensure safe and reliable operation. Among the different applications, obstacle identification is a primary module of the perception system. We propose a vision-based method built upon a deep convolutional neural network that can reason simultaneously about the location of objects in the image and their orientations on the ground plane. The same set of convolutional layers is used for the different tasks involved, avoiding the repetition of computations over the same image. Experiments on the KITTI dataset show that our efficiency-oriented method achieves state-of-the-art accuracies for object detection and viewpoint estimation, and is particularly suitable for the recognition of traffic situations from on-board vision systems. Code is available at https://github.com/cguindel/lsi-faster-rcnn.

Context-based object viewpoint estimation: A 2D relational approach

The task of object viewpoint estimation has been a challenge since the early days of computer vision. To estimate the viewpoint (or pose) of an object, people have mostly looked at object intrinsic features, such as shape or appearance. Surprisingly, informative features provided by other, extrinsic elements in the scene, have so far mostly been ignored. At the same time, contextual cues have been proven to be of great benefit for related tasks such as object detection or action recognition. In this paper, we explore how information from other objects in the scene can be exploited for viewpoint estimation. In particular, we look at object configurations by following a relational neighbor-based approach for reasoning about object relations. We show that, starting from noisy object detections and viewpoint estimates, exploiting the estimated viewpoint and location of other objects in the scene can lead to improved object viewpoint predictions. Experiments on the KITTI dataset demonstrate that object configurations can indeed be used as a complementary cue to appearance-based viewpoint estimation. Our analysis reveals that the proposed context-based method can improve object viewpoint estimation by reducing specific types of viewpoint estimation errors commonly made by methods that only consider local information. Moreover, considering contextual information produces superior performance in scenes where a high number of object instances occur. Finally, our results suggest that, following a cautious relational neighbor formulation brings improvements over its aggressive counterpart for the task of object viewpoint estimation.

Location-aware fine-grained vehicle type recognition using multi-task deep networks

Distinct from conventional approaches to generic vehicle type recognition, this paper proposes an approach to fine-grained vehicle type recognition that leverages multiple cues jointly to convolutional neural networks (CNNs). Fine-grained vehicle recognition is inherently difficult with many challenges, including (i) how to incorporate multiple cues through joint optimization instead of separately handling, (ii) how to learn data-driven representation for vehicles instead of hand-crafted features and (iii) how to perceive and localize the vehicle region instead of using the whole image. Our multi-task CNNs localize vehicles in the first stage and recognize subclasses in the second stage, allowing us to handle samples with cluttered backgrounds and those where vehicles do not fill most of the image. Through collaborative feature learning from multiple tasks in each stage, our approach can handle subtle inter-class variations at the subordinate level. We also develop new vehicle-oriented data augmentation strategies in CNN training. To advance research on vehicle-centered tasks, we release a new vehicle dataset that provides both semantic labels and bounding boxes. This dataset can be used in vehicle localization, recognition, viewpoint estimation, and so on. Extensive experiments on two benchmark vehicle datasets demonstrate that our approach outperforms state-of-the-art algorithms.

Robust Vehicle Detection and Viewpoint Estimation With Soft Discriminative Mixture Model

Vehicle detection and vehicle viewpoint estimation are both crucial for assistive and autonomous driving systems. In this paper, we propose a soft discriminative mixture of viewpoint (SDMoV) models for joint vehicle detection and vehicle viewpoint estimation. The proposed SDMoV model is learned in two steps. First, a discriminative viewpoint-specific component model, which aims to maximize vehicle viewpoint classification accuracy, is learned for each cluster of vehicle images with similar viewpoint. Second, a new soft margin objective function, which aims to maximize vehicle detection accuracy, is designed to retrain these component models into a mixture of viewpoint models. The proposed SDMoV model is capable of detecting vehicles and estimating their viewpoints simultaneously. Experiments on three state-of-the-art datasets show that the proposed SDMoV model achieves superior accuracy for both vehicle detection and vehicle viewpoint estimation tasks.