Felipe Codevilla

Universidade Federal do Rio Grande (FURG) | FURG · Center for Computer Science - C ³

On end-to-end driving, a large amount of expert driving demonstrations is used to train an agent that mimics the expert by predicting its control actions. This process is self-supervised on vehicle signals (e.g., steering angle, acceleration) and does not require extra costly supervision (human labeling). Yet, the improvement of existing self-super...

Humans have the innate ability to attend to the most relevant actors in their vicinity and can forecast how they may behave in the future. This ability will be crucial for the deployment of safety-critical agents such as robots or vehicles which interact with humans. We propose a theoretical framework for this problem setting based on autoregressiv...

Human drivers produce a vast amount of data which could, in principle, be used to improve autonomous driving systems. Unfortunately, seemingly straightforward approaches for creating end-to-end driving models that map sensor data directly into driving actions are problematic in terms of interpretability, and typically have significant difficulty de...

A crucial component of an autonomous vehicle (AV) is the artificial intelligence (AI) is able to drive towards a desired destination. Today, there are different paradigms addressing the development of AI drivers. On the one hand, we find modular pipelines, which divide the driving task into sub-tasks such as perception and maneuver planning and con...

Autonomous vehicles (AVs) are key for the intelligent mobility of the future. A crucial component of an AV is the artificial intelligence (AI) able to drive towards a desired destination. Today, there are different paradigms addressing the development of AI drivers. On the one hand, we find modular pipelines, which divide the driving task into sub-...

Driving requires reacting to a wide variety of complex environment conditions and agent behaviors. Explicitly modeling each possible scenario is unrealistic. In contrast, imitation learning can, in theory, leverage data from large fleets of human-driven cars. Behavior cloning in particular has been successfully used to learn simple visuomotor polic...

Autonomous driving models should ideally be evaluated by deploying them on a fleet of physical vehicles in the real world. Unfortunately, this approach is not practical for the vast majority of researchers. An attractive alternative is to evaluate models offline, on a pre-collected validation dataset with ground truth annotation. In this paper, we...

Autonomous driving models should ideally be evaluated by deploying them on a fleet of physical vehicles in the real world. Unfortunately, this approach is not practical for the vast majority of researchers. An attractive alternative is to evaluate models offline, on a pre-collected validation dataset with ground truth annotation. In this paper, we...

We introduce CARLA, an open-source simulator for autonomous driving research. CARLA has been developed from the ground up to support development, training, and validation of autonomous urban driving systems. In addition to open-source code and protocols, CARLA provides open digital assets (urban layouts, buildings, vehicles) that were created for t...

Deep networks trained on demonstrations of human driving have learned to follow roads and avoid obstacles. However, driving policies trained via imitation learning cannot be controlled at test time. A vehicle trained end-to-end to imitate an expert cannot be guided to take a specific turn at an upcoming intersection. This limits the utility of such...

Context information is fundamental for image understanding. Many algorithms add context information by including semantic relations among objects such as neighboring tendencies, relative sizes and positions. To achieve context inclusion, popular context-aware classification methods rely on probabilistic graphical models such as Markov Random Fields...

Methods to detect local features have been made to be invariant to many transformations. So far, the vast majority of feature detectors consider robustness just to over-land effects. However, when capturing pictures in underwater environments, there are media specific properties that can degrade the visual quality the captured images. Little work h...

The underwater vision is highly spoiled by the underwater degradation effects. As light propagates in the water or other participative mediums, it suffers from a substantial scattering effect that produces poor image quality. Based on a physical model that describes this phenomenon it is possible to recover an haze-free image. But, for this procedu...

The underwater vision is highly spoiled by the underwater degradation effects. As light propagates in the water or other participative mediums, it suffers from a substantial scattering effect that produces poor image quality. Based on a physical model that describes this phenomenon it is possible to recover an haze-free image. But, for this procedu...

This paper proposes a methodology to estimate the transmission in underwater environments which consists on an adaptation of the Dark Channel Prior (DCP), a statistical prior based on properties of images obtained in outdoor natural scenes. Our methodology, called Underwater DCP (UDCP), basically considers that the blue and green color channels are...

A common neural network used for complex data clustering is the Self Organizing Maps(SOM). This algorithm have a expensive training step, that occur mainly on high dimensional applications like image clustering. This makes impossible for some of these applications to be run in real time or even in a feasible time. On this paper we explore the use o...

The present paper describes the FURGBOL robot F-180 team. The FURGBOL RoboCup team uses inexpensive and easily extendible hardware components and a standard linux software environment. We propose a modular archi-tecture, having three main stages: i. a Deliberative Stage (associated with strategy and path planning issues), ii. a Com-munication Stage...