SVG-Net: An SVG-based Trajectory Prediction Model

Abstract

Anticipating motions of vehicles in a scene is an essential problem for safe autonomous driving systems. To this end, the comprehension of the scene's infrastructure is often the main clue for predicting future trajectories. Most of the proposed approaches represent the scene with a rasterized format and some of the more recent approaches leverage custom vectorized formats. In contrast, we propose representing the scene's information by employing Scalable Vector Graphics (SVG). SVG is a well-established format that matches the problem of trajectory prediction better than rasterized formats while being more general than arbitrary vectorized formats. SVG has the potential to provide the convenience and generality of raster-based solutions if coupled with a powerful tool such as CNNs, for which we introduce SVG-Net. SVG-Net is a Transformer-based Neural Network that can effectively capture the scene's information from SVG inputs. Thanks to the selfattention mechanism in its Transformers, SVG-Net can also adequately apprehend relations amongst the scene and the agents. We demonstrate SVG-Net's effectiveness by evaluating its performance on the publicly-available Argoverse forecasting dataset. Finally, we illustrate how, by using SVG, one can benefit from datasets and advancements in other research fronts that also utilize the same input format. Our code is available at https://vita-epfl.github.io/SVGNet/.

Vehicle trajectory prediction works, but not everywhere

“…The last column represents the results of the search method described in Section 3.3. We also reported the performance of considering only one category of scene generation functions in the optimization problem Equation (7). The results indicate a substantial increase in SOR and HOR across all baselines in different categories of the generated scenes.…”

“…Carnet [44] used attention mechanism to determine the scene regions that were attended more, leading to an interpretable solution. Some recent work showed that scene can be represented by vector format instead of images [7,24,32,46]. To further improve the reasoning of the model and generate predictions admissible with respect to the scene, use of symmetric cross-entropy loss [38], off-road loss [8], and REINFORCE loss [16] have been proposed.…”