Erickson R. Nascimento

Abdulrahman Kerim, Felipe Chamone, Washington Ramos et al.

Recent semantic segmentation models perform well under standard weather conditions and sufficient illumination but struggle with adverse weather conditions and nighttime. Collecting and annotating training data under these conditions is expensive, time-consuming, error-prone, and not always practical. Usually, synthetic data is used as a feasible data source to increase the amount of training data. However, just directly using synthetic data may actually harm the model's performance under normal weather conditions while getting only small gains in adverse situations. Therefore, we present a novel architecture specifically designed for using synthetic training data for domain adaptation. We propose a simple yet powerful addition to DeepLabV3+ by using weather and time-of-the-day supervisors trained with multi-task learning, making it both weather and nighttime aware, which improves its mIoU accuracy by $14$ percentage points on the ACDC dataset while maintaining a score of $75\%$ mIoU on the Cityscapes dataset. Our code is available at https://github.com/lsmcolab/Semantic-Segmentation-under-Adverse-Conditions.

Welerson Melo, Guilherme Potje, Felipe Cadar et al.

We present a novel learned keypoint detection method designed to maximize the number of correct matches for the task of non-rigid image correspondence. Our training framework uses true correspondences, obtained by matching annotated image pairs with a predefined descriptor extractor, as a ground-truth to train a convolutional neural network (CNN). We optimize the model architecture by applying known geometric transformations to images as the supervisory signal. Experiments show that our method outperforms the state-of-the-art keypoint detector on real images of non-rigid objects by 20 p.p. on Mean Matching Accuracy and also improves the matching performance of several descriptors when coupled with our detection method. We also employ the proposed method in one challenging realworld application: object retrieval, where our detector exhibits performance on par with the best available keypoint detectors. The source code and trained model are publicly available at https://github.com/verlab/LearningToDetect SIBGRAPI 2022

Guilherme Potje, Felipe Cadar, Andre Araujo et al.

Local feature extraction is a standard approach in computer vision for tackling important tasks such as image matching and retrieval. The core assumption of most methods is that images undergo affine transformations, disregarding more complicated effects such as non-rigid deformations. Furthermore, incipient works tailored for non-rigid correspondence still rely on keypoint detectors designed for rigid transformations, hindering performance due to the limitations of the detector. We propose DALF (Deformation-Aware Local Features), a novel deformation-aware network for jointly detecting and describing keypoints, to handle the challenging problem of matching deformable surfaces. All network components work cooperatively through a feature fusion approach that enforces the descriptors' distinctiveness and invariance. Experiments using real deforming objects showcase the superiority of our method, where it delivers 8% improvement in matching scores compared to the previous best results. Our approach also enhances the performance of two real-world applications: deformable object retrieval and non-rigid 3D surface registration. Code for training, inference, and applications are publicly available at https://verlab.dcc.ufmg.br/descriptors/dalf_cvpr23.

Guilherme Potje, Renato Martins, Felipe Cadar et al.

Most of the existing handcrafted and learning-based local descriptors are still at best approximately invariant to affine image transformations, often disregarding deformable surfaces. In this paper, we take one step further by proposing a new approach to compute descriptors from RGB-D images (where RGB refers to the pixel color brightness and D stands for depth information) that are invariant to isometric non-rigid deformations, as well as to scale changes and rotation. Our proposed description strategies are grounded on the key idea of learning feature representations on undistorted local image patches using surface geodesics. We design two complementary local descriptors strategies to compute geodesic-aware features efficiently: one efficient binary descriptor based on handcrafted binary tests (named GeoBit), and one learning-based descriptor (GeoPatch) with convolutional neural networks (CNNs) to compute features. In different experiments using real and publicly available RGB-D data benchmarks, they consistently outperforms state-of-the-art handcrafted and learning-based image and RGB-D descriptors in matching scores, as well as in object retrieval and non-rigid surface tracking experiments, with comparable processing times. We also provide to the community a new dataset with accurate matching annotations of RGB-D images of different objects (shirts, cloths, paintings, bags), subjected to strong non-rigid deformations, for evaluation benchmark of deformable surface correspondence algorithms.

Abdulrahman Kerim, Washington L. S. Ramos, Leandro Soriano Marcolino et al.

Video stabilization plays a central role to improve videos quality. However, despite the substantial progress made by these methods, they were, mainly, tested under standard weather and lighting conditions, and may perform poorly under adverse conditions. In this paper, we propose a synthetic-aware adverse weather robust algorithm for video stabilization that does not require real data and can be trained only on synthetic data. We also present Silver, a novel rendering engine to generate the required training data with an automatic ground-truth extraction procedure. Our approach uses our specially generated synthetic data for training an affine transformation matrix estimator avoiding the feature extraction issues faced by current methods. Additionally, since no video stabilization datasets under adverse conditions are available, we propose the novel VSAC105Real dataset for evaluation. We compare our method to five state-of-the-art video stabilization algorithms using two benchmarks. Our results show that current approaches perform poorly in at least one weather condition, and that, even training in a small dataset with synthetic data only, we achieve the best performance in terms of stability average score, distortion score, success rate, and average cropping ratio when considering all weather conditions. Hence, our video stabilization model generalizes well on real-world videos and does not require large-scale synthetic training data to converge.

Rafael Azevedo, Thiago Coutinho, João Ferreira et al.

Translating written sentences from oral languages to a sequence of manual and non-manual gestures plays a crucial role in building a more inclusive society for deaf and hard-of-hearing people. Facial expressions (non-manual), in particular, are responsible for encoding the grammar of the sentence to be spoken, applying punctuation, pronouns, or emphasizing signs. These non-manual gestures are closely related to the semantics of the sentence being spoken and also to the utterance of the speaker's emotions. However, most Sign Language Production (SLP) approaches are centered on synthesizing manual gestures and do not focus on modeling the speakers expression. This paper introduces a new method focused in synthesizing facial expressions for sign language. Our goal is to improve sign language production by integrating sentiment information in facial expression generation. The approach leverages a sentence sentiment and semantic features to sample from a meaningful representation space, integrating the bias of the non-manual components into the sign language production process. To evaluate our method, we extend the Frechet Gesture Distance (FGD) and propose a new metric called Frechet Expression Distance (FED) and apply an extensive set of metrics to assess the quality of specific regions of the face. The experimental results showed that our method achieved state of the art, being superior to the competitors on How2Sign and PHOENIX14T datasets. Moreover, our architecture is based on a carefully designed graph pyramid that makes it simpler, easier to train, and capable of leveraging emotions to produce facial expressions.

Guilherme Potje, Felipe Cadar, Andre Araujo et al.

We introduce a lightweight and accurate architecture for resource-efficient visual correspondence. Our method, dubbed XFeat (Accelerated Features), revisits fundamental design choices in convolutional neural networks for detecting, extracting, and matching local features. Our new model satisfies a critical need for fast and robust algorithms suitable to resource-limited devices. In particular, accurate image matching requires sufficiently large image resolutions - for this reason, we keep the resolution as large as possible while limiting the number of channels in the network. Besides, our model is designed to offer the choice of matching at the sparse or semi-dense levels, each of which may be more suitable for different downstream applications, such as visual navigation and augmented reality. Our model is the first to offer semi-dense matching efficiently, leveraging a novel match refinement module that relies on coarse local descriptors. XFeat is versatile and hardware-independent, surpassing current deep learning-based local features in speed (up to 5x faster) with comparable or better accuracy, proven in pose estimation and visual localization. We showcase it running in real-time on an inexpensive laptop CPU without specialized hardware optimizations. Code and weights are available at www.verlab.dcc.ufmg.br/descriptors/xfeat_cvpr24.

Abdulrahman Kerim, Leandro Soriano Marcolino, Erickson R. Nascimento et al.

Supervised machine learning methods require large-scale training datasets to perform well in practice. Synthetic data has been showing great progress recently and has been used as a complement to real data. However, there is yet a great urge to assess the usability of synthetically generated data. To this end, we propose a novel UCB-based training procedure combined with a dynamic usability metric. Our proposed metric integrates low-level and high-level information from synthetic images and their corresponding real and synthetic datasets, surpassing existing traditional metrics. By utilizing a UCB-based dynamic approach ensures continual enhancement of model learning. Unlike other approaches, our method effectively adapts to changes in the machine learning model's state and considers the evolving utility of training samples during the training process. We show that our metric is an effective way to rank synthetic images based on their usability. Furthermore, we propose a new attribute-aware bandit pipeline for generating synthetic data by integrating a Large Language Model with Stable Diffusion. Quantitative results show that our approach can boost the performance of a wide range of supervised classifiers. Notably, we observed an improvement of up to 10% in classification accuracy compared to traditional approaches, demonstrating the effectiveness of our approach. Our source code, datasets, and additional materials are publically available at https://github.com/A-Kerim/Synthetic-Data-Usability-2024.

Felipe Cadar, Welerson Melo, Vaishnavi Kanagasabapathi et al.

We propose a novel learned keypoint detection method to increase the number of correct matches for the task of non-rigid image correspondence. By leveraging true correspondences acquired by matching annotated image pairs with a specified descriptor extractor, we train an end-to-end convolutional neural network (CNN) to find keypoint locations that are more appropriate to the considered descriptor. For that, we apply geometric and photometric warpings to images to generate a supervisory signal, allowing the optimization of the detector. Experiments demonstrate that our method enhances the Mean Matching Accuracy of numerous descriptors when used in conjunction with our detection method, while outperforming the state-of-the-art keypoint detectors on real images of non-rigid objects by 20 p.p. We also apply our method on the complex real-world task of object retrieval where our detector performs on par with the finest keypoint detectors currently available for this task. The source code and trained models are publicly available at https://github.com/verlab/LearningToDetect_PRL_2023

Thiago L. Gomes, Thiago M. Coutinho, Rafael Azevedo et al.

This paper proposes a new end-to-end neural rendering architecture to transfer appearance and reenact human actors. Our method leverages a carefully designed graph convolutional network (GCN) to model the human body manifold structure, jointly with differentiable rendering, to synthesize new videos of people in different contexts from where they were initially recorded. Unlike recent appearance transferring methods, our approach can reconstruct a fully controllable 3D texture-mapped model of a person, while taking into account the manifold structure from body shape and texture appearance in the view synthesis. Specifically, our approach models mesh deformations with a three-stage GCN trained in a self-supervised manner on rendered silhouettes of the human body. It also infers texture appearance with a convolutional network in the texture domain, which is trained in an adversarial regime to reconstruct human texture from rendered images of actors in different poses. Experiments on different videos show that our method successfully infers specific body deformations and avoid creating texture artifacts while achieving the best values for appearance in terms of Structural Similarity (SSIM), Learned Perceptual Image Patch Similarity (LPIPS), Mean Squared Error (MSE), and Fréchet Video Distance (FVD). By taking advantages of both differentiable rendering and the 3D parametric model, our method is fully controllable, which allows controlling the human synthesis from both pose and rendering parameters. The source code is available at https://www.verlab.dcc.ufmg.br/retargeting-motion/wacv2022.

Felipe Cadar, Guilherme Potje, Renato Martins et al.

Visual correspondence is a crucial step in key computer vision tasks, including camera localization, image registration, and structure from motion. The most effective techniques for matching keypoints currently involve using learned sparse or dense matchers, which need pairs of images. These neural networks have a good general understanding of features from both images, but they often struggle to match points from different semantic areas. This paper presents a new method that uses semantic cues from foundation vision model features (like DINOv2) to enhance local feature matching by incorporating semantic reasoning into existing descriptors. Therefore, the learned descriptors do not require image pairs at inference time, allowing feature caching and fast matching using similarity search, unlike learned matchers. We present adapted versions of six existing descriptors, with an average increase in performance of 29% in camera localization, with comparable accuracy to existing matchers as LightGlue and LoFTR in two existing benchmarks. Both code and trained models are available at https://www.verlab.dcc.ufmg.br/descriptors/reasoning_accv24

Guilherme Potje, Renato Martins, Felipe Cadar et al.

Despite the advances in extracting local features achieved by handcrafted and learning-based descriptors, they are still limited by the lack of invariance to non-rigid transformations. In this paper, we present a new approach to compute features from still images that are robust to non-rigid deformations to circumvent the problem of matching deformable surfaces and objects. Our deformation-aware local descriptor, named DEAL, leverages a polar sampling and a spatial transformer warping to provide invariance to rotation, scale, and image deformations. We train the model architecture end-to-end by applying isometric non-rigid deformations to objects in a simulated environment as guidance to provide highly discriminative local features. The experiments show that our method outperforms state-of-the-art handcrafted, learning-based image, and RGB-D descriptors in different datasets with both real and realistic synthetic deformable objects in still images. The source code and trained model of the descriptor are publicly available at https://www.verlab.dcc.ufmg.br/descriptors/neurips2021.

Omar Vidal Pino, Erickson Rangel Nascimento, Mario Fernando Montenegro Campos

In this paper, we hypothesize that the effects of the degree of typicality in natural semantic categories can be generated based on the structure of artificial categories learned with deep learning models. Motivated by the human approach to representing natural semantic categories and based on the Prototype Theory foundations, we propose a novel Computational Prototype Model (CPM) to represent the internal structure of semantic categories. Unlike other prototype learning approaches, our mathematical framework proposes a first approach to provide deep neural networks with the ability to model abstract semantic concepts such as category central semantic meaning, typicality degree of an object's image, and family resemblance relationship. We proposed several methodologies based on the typicality's concept to evaluate our CPM-model in image semantic processing tasks such as image classification, a global semantic description, and transfer learning. Our experiments on different image datasets, such as ImageNet and Coco, showed that our approach might be an admissible proposition in the effort to endow machines with greater power of abstraction for the semantic representation of objects' categories.

Thiago L. Gomes, Renato Martins, João Ferreira et al.

Transferring human motion and appearance between videos of human actors remains one of the key challenges in Computer Vision. Despite the advances from recent image-to-image translation approaches, there are several transferring contexts where most end-to-end learning-based retargeting methods still perform poorly. Transferring human appearance from one actor to another is only ensured when a strict setup has been complied, which is generally built considering their training regime's specificities. In this work, we propose a shape-aware approach based on a hybrid image-based rendering technique that exhibits competitive visual retargeting quality compared to state-of-the-art neural rendering approaches. The formulation leverages the user body shape into the retargeting while considering physical constraints of the motion in 3D and the 2D image domain. We also present a new video retargeting benchmark dataset composed of different videos with annotated human motions to evaluate the task of synthesizing people's videos, which can be used as a common base to improve tracking the progress in the field. The dataset and its evaluation protocols are designed to evaluate retargeting methods in more general and challenging conditions. Our method is validated in several experiments, comprising publicly available videos of actors with different shapes, motion types, and camera setups. The dataset and retargeting code are publicly available to the community at: https://www.verlab.dcc.ufmg.br/retargeting-motion.

João P. Ferreira, Thiago M. Coutinho, Thiago L. Gomes et al.

Synthesizing human motion through learning techniques is becoming an increasingly popular approach to alleviating the requirement of new data capture to produce animations. Learning to move naturally from music, i.e., to dance, is one of the more complex motions humans often perform effortlessly. Each dance movement is unique, yet such movements maintain the core characteristics of the dance style. Most approaches addressing this problem with classical convolutional and recursive neural models undergo training and variability issues due to the non-Euclidean geometry of the motion manifold structure.In this paper, we design a novel method based on graph convolutional networks to tackle the problem of automatic dance generation from audio information. Our method uses an adversarial learning scheme conditioned on the input music audios to create natural motions preserving the key movements of different music styles. We evaluate our method with three quantitative metrics of generative methods and a user study. The results suggest that the proposed GCN model outperforms the state-of-the-art dance generation method conditioned on music in different experiments. Moreover, our graph-convolutional approach is simpler, easier to be trained, and capable of generating more realistic motion styles regarding qualitative and different quantitative metrics. It also presented a visual movement perceptual quality comparable to real motion data.

Alan Carvalho Neves, Michel Melo Silva, Mario Fernando Montenegro Campos et al.

The growing data sharing and life-logging cultures are driving an unprecedented increase in the amount of unedited First-Person Videos. In this paper, we address the problem of accessing relevant information in First-Person Videos by creating an accelerated version of the input video and emphasizing the important moments to the recorder. Our method is based on an attention model driven by gaze and visual scene analysis that provides a semantic score of each frame of the input video. We performed several experimental evaluations on publicly available First-Person Videos datasets. The results show that our methodology can fast-forward videos emphasizing moments when the recorder visually interact with scene components while not including monotonous clips.

Washington Ramos, Michel Silva, Edson Araujo et al.

The rapid increase in the amount of published visual data and the limited time of users bring the demand for processing untrimmed videos to produce shorter versions that convey the same information. Despite the remarkable progress that has been made by summarization methods, most of them can only select a few frames or skims, which creates visual gaps and breaks the video context. In this paper, we present a novel methodology based on a reinforcement learning formulation to accelerate instructional videos. Our approach can adaptively select frames that are not relevant to convey the information without creating gaps in the final video. Our agent is textually and visually oriented to select which frames to remove to shrink the input video. Additionally, we propose a novel network, called Visually-guided Document Attention Network (VDAN), able to generate a highly discriminative embedding space to represent both textual and visual data. Our experiments show that our method achieves the best performance in terms of F1 Score and coverage at the video segment level.

Renato Martins, Dhiego Bersan, Mario F. M. Campos et al.

This paper addresses the problem of building augmented metric representations of scenes with semantic information from RGB-D images. We propose a complete framework to create an enhanced map representation of the environment with object-level information to be used in several applications such as human-robot interaction, assistive robotics, visual navigation, or in manipulation tasks. Our formulation leverages a CNN-based object detector (Yolo) with a 3D model-based segmentation technique to perform instance semantic segmentation, and to localize, identify, and track different classes of objects in the scene. The tracking and positioning of semantic classes is done with a dictionary of Kalman filters in order to combine sensor measurements over time and then providing more accurate maps. The formulation is designed to identify and to disregard dynamic objects in order to obtain a medium-term invariant map representation. The proposed method was evaluated with collected and publicly available RGB-D data sequences acquired in different indoor scenes. Experimental results show the potential of the technique to produce augmented semantic maps containing several objects (notably doors). We also provide to the community a dataset composed of annotated object classes (doors, fire extinguishers, benches, water fountains) and their positioning, as well as the source code as ROS packages.

Thiago L. Gomes, Renato Martins, João Ferreira et al.

Creating plausible virtual actors from images of real actors remains one of the key challenges in computer vision and computer graphics. Marker-less human motion estimation and shape modeling from images in the wild bring this challenge to the fore. Although the recent advances on view synthesis and image-to-image translation, currently available formulations are limited to transfer solely style and do not take into account the character's motion and shape, which are by nature intermingled to produce plausible human forms. In this paper, we propose a unifying formulation for transferring appearance and retargeting human motion from monocular videos that regards all these aspects. Our method synthesizes new videos of people in a different context where they were initially recorded. Differently from recent appearance transferring methods, our approach takes into account body shape, appearance, and motion constraints. The evaluation is performed with several experiments using publicly available real videos containing hard conditions. Our method is able to transfer both human motion and appearance outperforming state-of-the-art methods, while preserving specific features of the motion that must be maintained (e.g., feet touching the floor, hands touching a particular object) and holding the best visual quality and appearance metrics such as Structural Similarity (SSIM) and Learned Perceptual Image Patch Similarity (LPIPS).

Washington L. S. Ramos, Michel M. Silva, Edson R. Araujo et al.

The growth of Social Networks has fueled the habit of people logging their day-to-day activities, and long First-Person Videos (FPVs) are one of the main tools in this new habit. Semantic-aware fast-forward methods are able to decrease the watch time and select meaningful moments, which is key to increase the chances of these videos being watched. However, these methods can not handle semantics in terms of personalization. In this work, we present a new approach to automatically creating personalized fast-forward videos for FPVs. Our approach explores the availability of text-centric data from the user's social networks such as status updates to infer her/his topics of interest and assigns scores to the input frames according to her/his preferences. Extensive experiments are conducted on three different datasets with simulated and real-world users as input, achieving an average F1 score of up to 12.8 percentage points higher than the best competitors. We also present a user study to demonstrate the effectiveness of our method.

Omar Vidal Pino, Erickson Rangel Nascimento, Mario Fernando Montenegro Campos

In this paper, we introduce a novel semantic description approach inspired on Prototype Theory foundations. We propose a Computational Prototype Model (CPM) that encodes and stores the central semantic meaning of objects category: the semantic prototype. Also, we introduce a Prototype-based Description Model that encodes the semantic meaning of an object while describing its features using our CPM model. Our description method uses semantic prototypes computed by CNN-classifications models to create discriminative signatures that describe an object highlighting its most distinctive features within the category. Our experiments show that: i) our CPM model (semantic prototype + distance metric) is able to describe the internal semantic structure of objects categories; ii) our semantic distance metric can be understood as the object visual typicality score within a category; iii) our descriptor encoding is semantically interpretable and significantly outperforms other image global encodings in clustering and classification tasks.

Bruna Vieira Frade, Erickson R. Nascimento

The detection of fiducial points on faces has significantly been favored by the rapid progress in the field of machine learning, in particular in the convolution networks. However, the accuracy of most of the detectors strongly depends on an enormous amount of annotated data. In this work, we present a domain adaptation approach based on a two-step learning to detect fiducial points on human and animal faces. We evaluate our method on three different datasets composed of different animal faces (cats, dogs, and horses). The experiments show that our method performs better than state of the art and can use few annotated data to leverage the detection of landmarks reducing the demand for large volume of annotated data.

Vinicius S. Furlan, Ruzena Bajcsy, Erickson R. Nascimento

The remarkable technological advance in well-equipped wearable devices is pushing an increasing production of long first-person videos. However, since most of these videos have long and tedious parts, they are forgotten or never seen. Despite a large number of techniques proposed to fast-forward these videos by highlighting relevant moments, most of them are image based only. Most of these techniques disregard other relevant sensors present in the current devices such as high-definition microphones. In this work, we propose a new approach to fast-forward videos using psychoacoustic metrics extracted from the soundtrack. These metrics can be used to estimate the annoyance of a segment allowing our method to emphasize moments of sound pleasantness. The efficiency of our method is demonstrated through qualitative results and quantitative results as far as of speed-up and instability are concerned.

Michel Melo Silva, Washington Luis Souza Ramos, Felipe Cadar Chamone et al.

The emergence of low-cost high-quality personal wearable cameras combined with the increasing storage capacity of video-sharing websites have evoked a growing interest in first-person videos, since most videos are composed of long-running unedited streams which are usually tedious and unpleasant to watch. State-of-the-art semantic fast-forward methods currently face the challenge of providing an adequate balance between smoothness in visual flow and the emphasis on the relevant parts. In this work, we present the Multi-Importance Fast-Forward (MIFF), a fully automatic methodology to fast-forward egocentric videos facing these challenges. The dilemma of defining what is the semantic information of a video is addressed by a learning process based on the preferences of the user. Results show that the proposed method keeps over $3$ times more semantic content than the state-of-the-art fast-forward. Finally, we discuss the need of a particular video stabilization technique for fast-forward egocentric videos.

Guilherme Nascimento, Camila Laranjeira, Vinicius Braz et al.

In this paper, we present a robust method for scene recognition, which leverages Convolutional Neural Networks (CNNs) features and Sparse Coding setting by creating a new representation of indoor scenes. Although CNNs highly benefited the fields of computer vision and pattern recognition, convolutional layers adjust weights on a global-approach, which might lead to losing important local details such as objects and small structures. Our proposed scene representation relies on both: global features that mostly refers to environment's structure, and local features that are sparsely combined to capture characteristics of common objects of a given scene. This new representation is based on fragments of the scene and leverages features extracted by CNNs. The experimental evaluation shows that the resulting representation outperforms previous scene recognition methods on Scene15 and MIT67 datasets, and performs competitively on SUN397, while being highly robust to perturbations in the input image such as noise and occlusion.

Manoel Horta Ribeiro, Bruno Teixeira, Antônio Otávio Fernandes et al.

Many of the state-of-the-art algorithms for gesture recognition are based on Conditional Random Fields (CRFs). Successful approaches, such as the Latent-Dynamic CRFs, extend the CRF by incorporating latent variables, whose values are mapped to the values of the labels. In this paper we propose a novel methodology to set the latent values according to the gesture complexity. We use an heuristic that iterates through the samples associated with each label value, stimating their complexity. We then use it to assign the latent values to the label values. We evaluate our method on the task of recognizing human gestures from video streams. The experiments were performed in binary datasets, generated by grouping different labels. Our results demonstrate that our approach outperforms the arbitrary one in many cases, increasing the accuracy by up to 10%.