Karlo Koledić

Tin Mišić, Karlo Koledić, Fabio Bonsignorio et al.

The ability to selectively attend to relevant stimuli while filtering out distractions is essential for agents that process complex, high-dimensional sensory input. This paper introduces a model of covert and overt visual attention through the framework of active inference, utilizing dynamic optimization of sensory precisions to minimize free-energy. The model determines visual sensory precisions based on both current environmental beliefs and sensory input, influencing attentional allocation in both covert and overt modalities. To test the effectiveness of the model, we analyze its behavior in the Posner cueing task and a simple target focus task using two-dimensional(2D) visual data. Reaction times are measured to investigate the interplay between exogenous and endogenous attention, as well as valid and invalid cueing. The results show that exogenous and valid cues generally lead to faster reaction times compared to endogenous and invalid cues. Furthermore, the model exhibits behavior similar to inhibition of return, where previously attended locations become suppressed after a specific cue-target onset asynchrony interval. Lastly, we investigate different aspects of overt attention and show that involuntary, reflexive saccades occur faster than intentional ones, but at the expense of adaptability.

Karlo Koledić, Luka Petrović, Ivan Marković et al.

Generalizing metric monocular depth estimation presents a significant challenge due to its ill-posed nature, while the entanglement between camera parameters and depth amplifies issues further, hindering multi-dataset training and zero-shot accuracy. This challenge is particularly evident in autonomous vehicles and mobile robotics, where data is collected with fixed camera setups, limiting the geometric diversity. Yet, this context also presents an opportunity: the fixed relationship between the camera and the ground plane imposes additional perspective geometry constraints, enabling depth regression via vertical image positions of objects. However, this cue is highly susceptible to overfitting, thus we propose a novel canonical representation that maintains consistency across varied camera setups, effectively disentangling depth from specific parameters and enhancing generalization across datasets. We also propose a novel architecture that adaptively and probabilistically fuses depths estimated via object size and vertical image position cues. A comprehensive evaluation demonstrates the effectiveness of the proposed approach on five autonomous driving datasets, achieving accurate metric depth estimation for varying resolutions, aspect ratios and camera setups. Notably, we achieve comparable accuracy to existing zero-shot methods, despite training on a single dataset with a single-camera setup. Project website: https://unizgfer-lamor.github.io/gvdepth/

Karlo Koledić, Luka Petrović, Ivan Petrović et al.

Learning-based monocular depth estimation leverages geometric priors present in the training data to enable metric depth perception from a single image, a traditionally ill-posed problem. However, these priors are often specific to a particular domain, leading to limited generalization performance on unseen data. Apart from the well studied environmental domain gap, monocular depth estimation is also sensitive to the domain gap induced by varying camera parameters, an aspect that is often overlooked in current state-of-the-art approaches. This issue is particularly evident in autonomous driving scenarios, where datasets are typically collected with a single vehicle-camera setup, leading to a bias in the training data due to a fixed perspective geometry. In this paper, we challenge this trend and introduce GenDepth, a novel model capable of performing metric depth estimation for arbitrary vehicle-camera setups. To address the lack of data with sufficiently diverse camera parameters, we first create a bespoke synthetic dataset collected with different vehicle-camera systems. Then, we design GenDepth to simultaneously optimize two objectives: (i) equivariance to the camera parameter variations on synthetic data, (ii) transferring the learned equivariance to real-world environmental features using a single real-world dataset with a fixed vehicle-camera system. To achieve this, we propose a novel embedding of camera parameters as the ground plane depth and present a novel architecture that integrates these embeddings with adversarial domain alignment. We validate GenDepth on several autonomous driving datasets, demonstrating its state-of-the-art generalization capability for different vehicle-camera systems.