Publications

Despite the progress made in the style transfer task, most previous work focus on transferring only relatively simple features like color or texture, while missing more abstract concepts such as overall art expression or painter-specific traits. However, these abstract semantics can be captured by models like DALL-E or CLIP, which have been trained using huge datasets of images and textual documents. In this paper, we propose StylerDALLE, a style transfer method that exploits both of these models and uses natural language to describe abstract art styles. Specifically, we formulate the language-guided style transfer task as a non-autoregressive token sequence translation, i.e., from input content image to output stylized image, in the discrete latent space of a large-scale pretrained vector-quantized tokenizer, e.g., the discrete variational auto-encoder (dVAE) of DALL-E. To incorporate style information, we propose a Reinforcement Learning strategy with CLIP-based language supervision that ensures stylization and content preservation simultaneously. Experimental results demonstrate the superiority of our method, which can effectively transfer art styles using language instructions at different granularities. Code is available at https://github.com/zipengxuc/StylerDALLE.

Image captioning is a challenging task that combines Computer Vision and Natural Language Processing to generate descriptive and accurate textual descriptions for input images. Research efforts in this field mainly focus on developing novel architectural components to extend image captioning models and using large-scale image-text datasets crawled from the web to boost final performance. In this work, we explore an alternative to web-crawled data and augment the training dataset with synthetic images generated by a latent diffusion model. In particular, we propose a simple yet effective synthetic data augmentation framework that is capable of significantly improving the quality of captions generated by a standard Transformer-based model, leading to competitive results on the COCO dataset.

Explainability and interpretability of deep neural networks have become of crucial importance over the years in Computer Vision, concurrently with the need to understand increasingly complex models. This necessity has fostered research on approaches that facilitate human comprehension of neural methods. In this work, we propose an explainable setting for visual navigation, in which an autonomous agent needs to explore an unseen indoor environment while portraying and explaining interesting scenes with natural language descriptions. We combine recent advances in ongoing research fields, employing an explainability method on images generated through agent-environment interaction. Our approach uses explainable maps to visualize model predictions and highlight the correlation between the observed entities and the generated words, to focus on prominent objects encountered during the environment exploration. The experimental section demonstrates that our approach can identify the regions of the images that the agent concentrates on to describe its point of view, improving explainability.

The field of multi-object tracking has recently seen a renewed interest in the good old schema of tracking-by-detection, as its simplicity and strong priors spare it from the complex design and painful babysitting of tracking-by-attention approaches. In view of this, we aim at extending tracking-by-detection to multi-modal settings, where a comprehensive cost has to be computed from heterogeneous information e.g., 2D motion cues, visual appearance, and pose estimates. More precisely, we follow a case study where a rough estimate of 3D information is also available and must be merged with other traditional metrics (e.g., the IoU). To achieve that, recent approaches resort to either simple rules or complex heuristics to balance the contribution of each cost. However, i) they require careful tuning of tailored hyperparameters on a hold-out set, and ii) they imply these costs to be independent, which does not hold in reality. We address these issues by building upon an elegant probabilistic formulation, which considers the cost of a candidate association as the negative log-likelihood yielded by a deep density estimator, trained to model the conditional joint probability distribution of correct associations. Our experiments, conducted on both simulated and real benchmarks, show that our approach consistently enhances the performance of several tracking-by-detection algorithms.

Melanoma is a malignant cancer type which develops when DNA damage occurs (mainly due to environmental factors such as ultraviolet rays). Often, melanoma results in intense and aggressive cell growth that, if not caught in time, can bring one toward death. Thus, early identification at the initial stage is fundamental to stopping the spread of cancer. In this paper, a ViT-based architecture able to classify melanoma versus non-cancerous lesions is presented. The proposed predictive model is trained and tested on public skin cancer data from the ISIC challenge, and the obtained results are highly promising. Different classifier configurations are considered and analyzed in order to find the most discriminating one. The best one reached an accuracy of 0.948, sensitivity of 0.928, specificity of 0.967, and AUROC of 0.948.

With the recent explosion of interest in visual Generative AI, the field of deepfake detection has gained a lot of attention. In fact, deepfake detection might be the only measure to counter the potential proliferation of generated media in support of fake news and its consequences. While many of the available works limit the detection to a pure and direct classification of fake versus real, this does not translate well to a real-world scenario. Indeed, malevolent users can easily apply post-processing techniques to generated content, changing the underlying distribution of fake data. In this work, we provide an in-depth analysis of the robustness of a deepfake detection pipeline, considering different image augmentations, transformations, and other pre-processing steps. These transformations are only applied in the evaluation phase, thus simulating a practical situation in which the detector is not trained on all the possible augmentations that can be used by the attacker. In particular, we analyze the performance of a k-NN and a linear probe detector on the COCOFake dataset, using image features extracted from pre-trained models, like CLIP and DINO. Our results demonstrate that while the CLIP visual backbone outperforms DINO in deepfake detection with no augmentation, its performance varies significantly in presence of any transformation, favoring the robustness of DINO.

A principled approach to the analysis of eye movements for behavioural biometrics is laid down. The approach grounds in foraging theory, which provides a sound basis to capture the unique- ness of individual eye movement behaviour. We propose a composite Ornstein-Uhlenbeck process for quantifying the exploration/exploitation signature characterising the foraging eye behaviour. The rel- evant parameters of the composite model, inferred from eye-tracking data via Bayesian analysis, are shown to yield a suitable feature set for biometric identification; the latter is eventually accomplished via a classical classification technique. A proof of concept of the method is provided by measuring its identification performance on a publicly available dataset. Data and code for reproducing the analyses are made available. Overall, we argue that the approach offers a fresh view on either the analyses of eye-tracking data and prospective applications in this field.

: The accuracy and the overall performances of ophthalmic instrumentation, where specific analysis of eye images is involved, can be negatively influenced by invalid or incorrect frames acquired during everyday measurements of unaware or non-collaborative human patients and non-technical operators. Therefore, in this paper, we investigate and compare the adoption of several vision-based classification algorithms belonging to different fields, i.e., Machine Learning, Deep Learning, and Expert Systems, in order to improve the performance of an ophthalmic instrument designed for the Pupillary Light Reflex measurement. To test the implemented solutions, we collected and publicly released PopEYE as one of the first datasets consisting of 15 k eye images belonging to 22 different subjects acquired through the aforementioned specialized ophthalmic device. Finally, we discuss the experimental results in terms of classification accuracy of the eye status, as well as computational load analysis, since the proposed solution is designed to be implemented in embedded boards, which have limited hardware resources in computational power and memory size.