Long-term trajectory forecasting is an important and challenging problem in the fields of computer vision, machine learning, and robotics. One fundamental difficulty stands in the evolution of the trajectory that becomes more and more uncertain and unpredictable as the time horizon grows, subsequently increasing the complexity of the problem. To overcome this issue, in this paper, we propose Di-Long, a new method that employs the distillation of a short-term trajectory model forecaster that guides a student network for long-term trajectory prediction during the training process. Given a total sequence length that comprehends the allowed observation for the student network and the complementary target sequence, we let the student and the teacher solve two different related tasks defined over the same full trajectory: the student observes a short sequence and predicts a long trajectory, whereas the teacher observes a longer sequence and predicts the remaining short target trajectory. The teacher’s task is less uncertain, and we use its accurate predictions to guide the student through our knowledge distillation framework, reducing long-term future uncertainty. Our experiments show that our proposed Di-Long method is effective for long-term forecasting and achieves state-of-the-art performance on the Intersection Drone Dataset (inD) and the Stanford Drone Dataset (SDD).

Language-vision integration has become an increasingly popular research direction within the computer vision field. In recent years, there has been a growing recognition of the importance of incorporating linguistic information into visual tasks, particularly in domains such as action anticipation. This integration allows anticipation models to leverage textual descriptions to gain deeper contextual understanding, leading to more accurate predictions. In this work, we focus on pedestrian action anticipation, where the objective is the early prediction of pedestrians’ future actions in urban environments. Our method relies on a multi-modal transformer model that encodes past observations and produces predictions at different anticipation times, employing a learned mask technique to filter out redundancy in the observed frames. Instead of relying solely on visual cues extracted from images or videos, we explore the impact of integrating textual information in enriching the input modalities of our pedestrian action anticipation model. We investigate various techniques for generating descriptive captions corresponding to input images, aiming to enhance the anticipation performance. Evaluation results on available public benchmarks demonstrate the effectiveness of our method in improving the prediction performance at different anticipation times compared to previous works. Additionally, incorporating the language modality in our anticipation model proved significant improvement, reaching a 29.5% increase in the F1 score at 1-second anticipation and a 16.66% increase at 4-second anticipation. These results underscore the potential of language-vision integration in advancing pedestrian action anticipation in complex urban environments.

Human intention prediction is a growing area of research where an activity in a video has to be anticipated by a vision-based system. To this end, the model creates a representation of the past, and subsequently, it produces future hypotheses about upcoming scenarios. In this work, we focus on pedestrians’ early intention prediction in which, from a current observation of an urban scene, the model predicts the future activity of pedestrians that approach the street. Our method is based on a multi-modal transformer that encodes past observations and produces multiple predictions at different anticipation times. Moreover, we propose to learn the attention masks of our transformer-based model (Temporal Adaptive Mask Transformer) in order to weigh differently present and past temporal dependencies. We investigate our method on several public benchmarks for early intention prediction, improving the prediction performances at different anticipation times compared to the previous works.

ReLU networks have remained the default choice for models in the area of image prediction despite their well-established spectral bias towards learning low frequencies faster, and consequently their difficulty of reproducing high frequency visual details. As an alternative, sin networks showed promising results in learning implicit representations of visual data. However training these networks in practically relevant settings proved to be difficult, requiring careful initialization, dealing with issues due to inconsistent gradients, and a degeneracy in local minima. In this work, we instead propose replacing a baseline network’s existing activations with a novel ensemble function with trainable parameters. The proposed MetaSin activation can be trained reliably without requiring intricate initialization schemes, and results in consistently lower test loss compared to alternatives. We demonstrate our method in the areas of Monte-Carlo denoising and image resampling where we set new state-of-the-art through a knowledge distillation based training procedure. We present ablations on hyper-parameter settings, comparisons with alternative activation function formulations, and discuss the use of our method in other domains, such as image classification.

Anticipating human motion is an essential requirement for autonomous vehicles and robots in order to primary guarantee people’s safety. In urban scenarios, they interact with humans, the surrounding environment, and other vehicles relying on several cues to forecast crossing or not crossing intentions. For these reasons, this challenging task is often tackled using both visual and non-visual features to anticipate future actions from 2 s to 1 s earlier the event. Our work primarily aims to revise this standard evaluation protocol to forecast crossing events as early as possible. To this end, we conceive a solution upon an extensively used model for egocentric action anticipation (RU-LSTM), proposing to envision future features, or modalities, that can better infer human intentions using a properly attention-based fusion mechanism. We validate our model against JAAD and PIE datasets and demonstrate that an intent prediction model can benefit from these additional clues for anticipating pedestrians crossing events.

Automatic Speech Recognition (ASR) based on Recurrent Neural Network Transducers (RNN-T) is gaining interest in the speech community. We investigate data selection and preparation choices aiming for improved robustness of RNN-T ASR to speech disfluencies with a focus on partial words. For evaluation we use clean data, data with disfluen- cies and a separate dataset with speech affected by stuttering. We show that after including a small amount of data with disfluencies in the training set the recognition accuracy on the tests with disfluencies and stuttering improves. Increasing the amount of training data with disfluencies gives additional gains without degradation on the clean data. We also show that replacing partial words with a dedicated token helps to get even better accuracy on utterances with disfluencies and stutter. The evaluation of our best model shows 22.5% and 16.4% relative WER reduction on those two evaluation sets.

In open set recognition, a classifier has to detect unknown classes that are not known at training time. In order to recognize new categories, the classifier has to project the input samples of known classes in very compact and separated regions of the features space for discriminating samples of unknown classes. Recently proposed Capsule Networks have shown to outperform alternatives in many fields, particularly in image recognition, however they have not been fully applied yet to open-set recognition. In capsule networks, scalar neurons are replaced by capsule vectors or matrices, whose entries represent different proper-ties of objects. In our proposal, during training, capsules features of the same known class are encouraged to match a pre-defined gaussian, one for each class. To this end, we use the variational autoencoder framework, with a set of gaussian priors as the approximation for the posterior distribution. In this way, we are able to control the compactness of the features of the same class around the center of the gaussians, thus controlling the ability of the classifier in detecting samples from unknown classes. We conducted several experiments and ablation of our model, obtaining state of the art results on different datasets in the open set recognition and unknown detection tasks.

Human capability to anticipate near future from visual observations and non-verbal cues is essential for developing intelligent systems that need to interact with people. Several research areas, such as human-robot interaction (HRI), assisted living or autonomous driving need to foresee future events to avoid crashes or help people. Egocentric scenarios are classic examples where action anticipation is applied due to their numerous applications. Such challenging task demands to capture and model domain's hidden structure to reduce prediction uncertainty. Since multiple actions may equally occur in the future, we treat action anticipation as a multi-label problem with missing labels extending the concept of label smoothing. This idea resembles the knowledge distillation process since useful information is injected into the model during training. We implement a multi-modal framework based on long short-term memory (LSTM) networks to summarize past observations and make predictions at different time steps. We perform extensive experiments on EPIC-Kitchens and EGTEA Gaze+ datasets including more than 2500 and 100 action classes, respectively. The experiments show that label smoothing systematically improves performance of state-of-the-art models for action anticipation.