Video Object Segmentation

With the widespread adoption of digital technologies, particularly mobile devices, the amount of available video data has skyrocketed. As a result, tools are urgently needed to facilitate automatic video analysis and comprehension. In this context, video object segmentation (VOS) is a crucial task, with potentially benefited areas ranging from video processing activities, like video compression and captioning, to a variety of applications, including visual tracking, video-based question answering, human pose estimation, surveillance and autonomous driving. The main goal of VOS is to classify all the pixels along a frame sequence into foreground and background regions.

Video summarization

Hours of video are uploaded to streaming platforms every minute, with recommender systems suggesting popular and relevant videos that can help users save time in the searching process. Recommender systems regularly require video summarization as an expert system to automatically identify suitable video entities and events. Since there is no well-established methodology to evaluate the relevance of summarized videos, some studies have made use of user annotations to gather evidence about the effectiveness of summarization methods.

Traffic analysis

Correctly identifying the road area on an image is a crucial task for many traffic analyses based on surveillance cameras and computer vision. Despite that, most of the systems do not provide this functionality in an automatic fashion; instead, the road area needs to be annotated by tedious and inefficient manual processes. This situation results in further inconveniences when one deals with a lot of cameras, demanding considerable effort to setup the system. Besides, since traffic analysis is an outdoor activity, cameras are exposed to disturbances due to natural events (e.g., wind, rain and bird strikes), which may require recurrent system reconfiguration.

Check out this background subtraction library at Github

Synthetic data sets

To generate any data set, three steps are required: locate 3D models in the scene, set the discretization parameters, and finally run the generator. The set of rendered images provide data that can be used for geometric pattern recognition problems, such as depth estimation, camera pose estimation, 3D box estimation, 3D reconstruction, camera calibration, and also a pixel-perfect ground-truth for scene understanding problems (e.g., semantic segmentation, instance segmentation, object detection, just to cite a few).

Camera auto-calibration

Surveillance cameras are commonly used in public and private security systems. This kind of equipment allows automation of surveillance tasks, when integrated with intelligent pattern recognition systems. Camera calibration allows intelligent systems to use the 3D geometry of a scene as a tool to determine the position and size of a target object. Typical systems may contain a large number of cameras, which are installed in different locations, and they are composed of static and heterogeneous cameras.

Person re-identification

The automatic person re-identification (re-id) problem resides in matching an unknown person image to a database of previously labeled images of people. Among several issues to cope with this research field, person re-id has to deal with person appearance and environment variations. As such, discriminative features to represent a person identity must be robust regardless those variations. Comparison among two image features is commonly accomplished by distance metrics. Although features and distance metrics can be handcrafted or trainable, the latter type has demonstrated more potential to breakthroughs in achieving state-of-the-art performance over public data sets. A recent paradigm that allows to work with trainable features is deep learning, which aims at learning features directly from raw image data. Although deep learning has recently achieved significant improvements in person re-identification, found on some few recent works, there is still room for learning strategies, which can be exploited to increase the current state-of-the-art performance.

Face recognition

Face is one of the most exploited human characteristics in automatic identification systems, being used both in access control security systems, or for identifying suspects, just to cite a few examples. Mainly in biometric access control systems, facial identification of a subject may be circumvented from photos, videos or masks, allowing for impostors to access such systems. These attacks occur by capturing an image of genuine face, recording eye or mouth movements of a user granted to access, or even masks made especially to mimic a genuine user of the system. Because of that, face spoofing detection methods become essential to support face recognition systems in order to distinguish a false image (image of an image or image of a video) from a real one (image of a user).

Person detection

Accuracy in image object detection has been usually achieved at the expense of much computational load. Therefore a trade-off between detection performance and fast execution commonly represents the ultimate goal of an object detector in real life applications. Most images are composed of non-trivial amounts of background information, such as sky, ground and water. In this sense, using an object detector against a recurring background pattern can require a significant amount of the total processing time. To alleviate this problem, search space reduction methods can help focusing the detection procedure on more distinctive image regions.

Gaze Estimation

Estimation of 3D gaze is highly relevant to multiple fields, including but not limited to interactive systems, specialized human-computer interfaces, and behavioral research. Although recently deep learning methods have boosted the accuracy of appearance based gaze estimation, there is still room for improvement in the network architectures for this particular task. Therefore we propose here a novel network architecture grounded on self-attention augmented convolutions to improve the quality of the learned features during the training of a shallower residual network. The rationale is that self-attention mechanism can help outperform deeper architectures by learning dependencies between distant regions in full-face images. This mechanism can also create better and more spatially-aware feature representations derived from the face and eye images before gaze regression. We dubbed our framework ARes-gaze, which explores our Attention-augmented ResNet (ARes-14) as twin convolutional backbones. In our experiments, results showed a decrease of the average angular error by 2.38% when compared to state-of-the-art methods on the MPIIFaceGaze data set, and a second-place on the EyeDiap data set. It is noteworthy that our proposed framework was the only one to reach high accuracy simultaneously on both data sets.

Pathospotter

The PathoSpotter project emerged through the desire and enthusiasm of the pathologist Dr. Washington Luís, from the Oswald Cruz Foundation in Brazil, for improving the clinical-pathological diagnoses for renal diseases. In 2014, Dr. Washington met Dr. Angelo Duarte from State University of Feira de Santana (Brazil) and both start the building of the project PathoSpotter. In the beginning, Pathospotter aimed to assist pathologists in the identification of glomerular lesions from kidney biopsies. As the work evolved, the aim shifted to make large-scale correlations of lesions with clinical and historical data of patients, to enlarge the knowledge in kidney pathology. Currently, the project Pathospotter intends to offer computational tools to: (i) Facilitate the consensus of clinicians and pathologists, helping to achieve more accurate and faster diagnoses; (ii) Facilitate large-scale clinical-pathological correlations studies; (iii) Help pathologists and students in classifying lesions in kidney biopsies. Check our site here.

OdontoAI

Segmenting teeth in dental X-ray images has been pursuit for many years, mainly relying in unsupervised methods. Although many approaches were proposed and tested, successful results were still far from being reached. Segmenting tooth in buccal images are mandatory for more complex tasks in decision support systems. This is the first step to detect not only teeth and their constituent parts, but also artifacts (e.g., prosthesis), tooth problems, and even missing teeth.

Physiotherapy

When the Physiotherapist meets the patient to start the process of evaluation and treatment, the movement analysis is carried out by considering each component, separately (joint or plan), or observing all aspects in an integrated perspective. Health professionals generally use clinical tests, such as rating scales, which can be less comprehensive andrequire subjective input. Other tools are also used, suchas: Force platforms, wearable devices, optoelectronic systems, electrogoniometers , electromyography and accelerometers. These instruments and clinical tests may not establish a true picture of the patient functional situation, because they can not accurate, objective and quantitatively measure whether the performance of activities is being im-proved or not. Thus, emerges the importance of developing new instruments and methods to assess functional movementin these patients.

Drones

Nowadays, a drone equipped with cameras is essential to explore remote areas. Drones capable of automatically performing low-altitude flights in places, such as roofed archaeological parks, caves, or sites covered by dense vegetation, are of great importance to guarantee a certain level of monitoring in difficult-to-access areas.

Media repercussion:

• Drones as Detectives: Surveying Crime Scenes for Evidence
• A New Method in Brazil Uses Drones as Crime Scene Investigators
• Drone detectives to identify evidence and create immersive simulations of crime scenes
• Criminal expert hopes to deploy drone detective

Sonar

To perceive the underwater environment, AUVs are commonly equipped with optical and sonar sensors. Optical sensors have limited sensing capability, i.e., they have a limited range of operation and require clear water conditions. On the other hand, sonars are better able to perceive the underwater environment, because they have a larger operational range even in turbid water conditions.

Thermal images

The discovery of infrared radiation and the development of technologies capable of detecting it made possible the emergence of thermography as a science. Thermography is a technique for graphically recording body temperatures, with the intention of distinguishing areas with different temperatures. Bodies with temperatures above -273◦C are capable of emitting infrared radiation.This characteristic allows to study the behavior of the temperature in different objects, structures and surfaces over time.

OxeBots

Considering strategies to promote robotics at the Federal University of Bahia, students from various engineering disciplines have come together to form a robotics team. The team has chosen the name “OxeBots,” which reflects the pride of representing Bahia through a distinctive expression. OxeBots is a team envisioned to go beyond competitions. We foster the study and application of concepts related to robotics, enhancing the technical and interpersonal skills of students and future professionals. The team consists of students from the Control and Automation Engineering, Computer Engineering, Mechanical Engineering, and Computer Science programs at the Federal University of Bahia. The team is affiliated with the IvisionLab. The main competition that the team will participate in is the annual RoboCup.