Denis F. Wolf

Angelica Tiemi Mizuno Nakamura, Denis Fernando Wolf, Valdir Grassi

Multi-Task Learning is a learning paradigm that uses correlated tasks to improve performance generalization. A common way to learn multiple tasks is through the hard parameter sharing approach, in which a single architecture is used to share the same subset of parameters, creating an inductive bias between them during the training process. Due to its simplicity, potential to improve generalization, and reduce computational cost, it has gained the attention of the scientific and industrial communities. However, tasks often conflict with each other, which makes it challenging to define how the gradients of multiple tasks should be combined to allow simultaneous learning. To address this problem, we use the idea of multi-objective optimization to propose a method that takes into account temporal behaviour of the gradients to create a dynamic bias that adjust the importance of each task during the backpropagation. The result of this method is to give more attention to the tasks that are diverging or that are not being benefited during the last iterations, allowing to ensure that the simultaneous learning is heading to the performance maximization of all tasks. As a result, we empirically show that the proposed method outperforms the state-of-art approaches on learning conflicting tasks. Unlike the adopted baselines, our method ensures that all tasks reach good generalization performances.

Iago Pachêco Gomes, Denis Fernando Wolf

Driving styles summarize different driving behaviors that reflect in the movements of the vehicles. These behaviors may indicate a tendency to perform riskier maneuvers, consume more fuel or energy, break traffic rules, or drive carefully. Therefore, this paper presents a driving style recognition using Interval Type-2 Fuzzy Inference System with Multiple Experts Decision-Making for classifying drivers into calm, moderate and aggressive. This system receives as input features longitudinal and lateral kinematic parameters of the vehicle motion. The type-2 fuzzy sets are more robust than type-1 fuzzy sets when handling noisy data, because their membership function are also fuzzy sets. In addition, a multiple experts approach can reduce the bias and imprecision while building the fuzzy rulebase, which stores the knowledge of the fuzzy system. The proposed approach was evaluated using descriptive statistics analysis, and compared with clustering algorithms and a type-1 fuzzy inference system. The results show the tendency to associate lower kinematic profiles for the driving styles classified with the type-2 fuzzy inference system when compared to other algorithms, which is in line with the more conservative approach adopted in the aggregation of the experts' opinions.

Luis Alberto Rosero, Iago Pacheco Gomes, Júnior Anderson Rodrigues da Silva et al.

The objective of the first CARLA autonomous driving challenge was to deploy autonomous driving systems to lead with complex traffic scenarios where all participants faced the same challenging traffic situations. According to the organizers, this competition emerges as a way to democratize and to accelerate the research and development of autonomous vehicles around the world using the CARLA simulator contributing to the development of the autonomous vehicle area. Therefore, this paper presents the architecture design for the navigation of an autonomous vehicle in a simulated urban environment that attempts to commit the least number of traffic infractions, which used as the baseline the original architecture of the platform for autonomous navigation CaRINA 2. Our agent traveled in simulated scenarios for several hours, demonstrating his capabilities, winning three out of the four tracks of the challenge, and being ranked second in the remaining track. Our architecture was made towards meeting the requirements of CARLA Autonomous Driving Challenge and has components for obstacle detection using 3D point clouds, traffic signs detection and classification which employs Convolutional Neural Networks (CNN) and depth information, risk assessment with collision detection using short-term motion prediction, decision-making with Markov Decision Process (MDP), and control using Model Predictive Control (MPC).

Thomio Watanabe, Denis F. Wolf

Deep Convolutional Neural Networks are able to identify complex patterns and perform tasks with super-human capabilities. However, besides the exceptional results, they are not completely understood and it is still impractical to hand-engineer similar solutions. In this work, an image classification Convolutional Neural Network and its building blocks are described from a frequency domain perspective. Some network layers have established counterparts in the frequency domain like the convolutional and pooling layers. We propose the 2SReLU layer, a novel non-linear activation function that preserves high frequency components in deep networks. It is demonstrated that in the frequency domain it is possible to achieve competitive results without using the computationally costly convolution operation. A source code implementation in PyTorch is provided at: https://gitlab.com/thomio/2srelu

Lucas P. N. Matias, Jefferson R. Souza, Denis F. Wolf

Robust perception systems are essential for autonomous vehicle safety. To navigate in a complex urban environment, it is necessary precise sensors with reliable data. The task of understanding the surroundings is hard by itself; for intelligent vehicles, it is even more critical due to the high speed in which the vehicle navigates. To successfully navigate in an urban environment, the perception system must quickly receive, process, and execute an action to guarantee both passenger and pedestrian safety. Stereo cameras collect environment information at many levels, e.g., depth, color, texture, shape, which guarantee ample knowledge about the surroundings. Even so, when compared to human, computational methods lack the ability to deal with missing information, i.e., occlusions. For many perception tasks, this lack of data can be a hindrance due to the environment incomplete information. In this paper, we address this problem and discuss recent methods to deal with occluded areas inference. We then introduce a loss function focused on disparity and environment depth data reconstruction, and a Generative Adversarial Network (GAN) architecture able to deal with occluded information inference. Our results present a coherent reconstruction on depth maps, estimating regions occluded by different obstacles. Our final contribution is a loss function focused on disparity data and a GAN able to extract depth features and estimate depth data by inpainting disparity images.

Daniela Ridel, Nachiket Deo, Denis Wolf et al.

Forecasting long-term human motion is a challenging task due to the non-linearity, multi-modality and inherent uncertainty in future trajectories. The underlying scene and past motion of agents can provide useful cues to predict their future motion. However, the heterogeneity of the two inputs poses a challenge for learning a joint representation of the scene and past trajectories. To address this challenge, we propose a model based on grid representations to forecast agent trajectories. We represent the past trajectories of agents using binary 2-D grids, and the underlying scene as a RGB birds-eye view (BEV) image, with an agent-centric frame of reference. We encode the scene and past trajectories using convolutional layers and generate trajectory forecasts using a Convolutional LSTM (ConvLSTM) decoder. Results on the publicly available Stanford Drone Dataset (SDD) show that our model outperforms prior approaches and outputs realistic future trajectories that comply with scene structure and past motion.

Daniela A. Ridel, Nachiket Deo, Denis Wolf et al.

Pedestrians and vehicles often share the road in complex inner city traffic. This leads to interactions between the vehicle and pedestrians, with each affecting the other's motion. In order to create robust methods to reason about pedestrian behavior and to design interfaces of communication between self-driving cars and pedestrians we need to better understand such interactions. In this paper, we present a data-driven approach to implicitly model pedestrians' interactions with vehicles, to better predict pedestrian behavior. We propose a LSTM model that takes as input the past trajectories of the pedestrian and ego-vehicle, and pedestrian head orientation, and predicts the future positions of the pedestrian. Our experiments based on a real-world, inner city dataset captured with vehicle mounted cameras, show that the usage of such cues improve pedestrian prediction when compared to a baseline that purely uses the past trajectory of the pedestrian.

Thomio Watanabe, Denis Wolf

The instance segmentation problem intends to precisely detect and delineate objects in images. Most of the current solutions rely on deep convolutional neural networks but despite this fact proposed solutions are very diverse. Some solutions approach the problem as a network problem, where they use several networks or specialize a single network to solve several tasks. A different approach tries to solve the problem as an annotation problem, where the instance information is encoded in a mathematical representation. This work proposes a solution based in the DCME technique to solve the instance segmentation with a single segmentation network. Different from others, the segmentation network decoder is not specialized in a multi-task network. Instead, the network encoder is repurposed to classify image objects, reducing the computational cost of the solution.

Thomio Watanabe, Denis Wolf

The autonomous car technology promises to replace human drivers with safer driving systems. But although autonomous cars can become safer than human drivers this is a long process that is going to be refined over time. Before these vehicles are deployed on urban roads a minimum safety level must be assured. Since the autonomous car technology is still under development there is no standard methodology to evaluate such systems. It is important to completely understand the technology that is being developed to design efficient means to evaluate it. In this paper we assume safety-critical systems reliability as a safety measure. We model an autonomous road vehicle as an intelligent agent and we approach its evaluation from an artificial intelligence perspective. Our focus is the evaluation of perception and decision making systems and also to propose a systematic method to evaluate their integration in the vehicle. We identify critical aspects of the data dependency from the artificial intelligence state of the art models and we also propose procedures to reproduce them.

Andrés E. Gómez, Tiago C. dos Santos, Carlos M. Massera et al.

According to data from the United Nations, more than 3000 people have died each day in the world due to road traffic collision. Considering recent researches, the human error may be considered as the main responsible for these fatalities. Because of this, researchers seek alternatives to transfer the vehicle control from people to autonomous systems. However, providing this technological innovation for the people may demand complex challenges in the legal, economic and technological areas. Consequently, carmakers and researchers have divided the driving automation in safety and emergency systems that improve the driver perception on the road. This may reduce the human error. Therefore, the main contribution of this study is to propose a driving simulator platform to develop and evaluate safety and emergency systems, in the first design stage. This driving simulator platform has an advantage: a flexible software structure.This allows in the simulation one adaptation for development or evaluation of a system. The proposed driving simulator platform was tested in two applications: cooperative vehicle system development and the influence evaluation of a Driving Assistance System (\textit{DAS}) on a driver. In the cooperative vehicle system development, the results obtained show that the increment of the time delay in the communication among vehicles ($V2V$) is determinant for the system performance. On the other hand, in the influence evaluation of a \textit{DAS} in a driver, it was possible to conclude that the \textit{DAS'} model does not have the level of influence necessary in a driver to avoid an accident.

Thomio Watanabe, Denis Wolf

The instance segmentation can be considered an extension of the object detection problem where bounding boxes are replaced by object contours. Strictly speaking the problem requires to identify each pixel instance and class independently of the artifice used for this mean. The advantage of instance segmentation over the usual object detection lies in the precise delineation of objects improving object localization. Additionally, object contours allow the evaluation of partial occlusion with basic image processing algorithms. This work approaches the instance segmentation problem as an annotation problem and presents a novel technique to encode and decode ground truth annotations. We propose a mathematical representation of instances that any deep semantic segmentation model can learn and generalize. Each individual instance is represented by a center of mass and a field of vectors pointing to it. This encoding technique has been denominated Distance to Center of Mass Encoding (DCME).

Caio César Teodoro Mendes, Vincent Frémont, Denis Fernando Wolf

Road detection is a fundamental task in autonomous navigation systems. In this paper, we consider the case of monocular road detection, where images are segmented into road and non-road regions. Our starting point is the well-known machine learning approach, in which a classifier is trained to distinguish road and non-road regions based on hand-labeled images. We proceed by introducing the use of "contextual blocks" as an efficient way of providing contextual information to the classifier. Overall, the proposed methodology, including its image feature selection and classifier, was conceived with computational cost in mind, leaving room for optimized implementations. Regarding experiments, we perform a sensible evaluation of each phase and feature subset that composes our system. The results show a great benefit from using contextual blocks and demonstrate their computational efficiency. Finally, we submit our results to the KITTI road detection benchmark achieving scores comparable with state of the art methods.