Abstract: Document Visual Question Answering (DocVQA) is a fast growing branch of document understanding. Despite the fact that documents contain sensitive or copyrighted information, none of the current DocVQA methods offers strong privacy guarantees.
In this work, we explore privacy in the domain of DocVQA for the first time. We highlight privacy issues in state of the art multi-modal LLM models used for DocVQA, and explore possible solutions.
Specifically, we focus on the invoice processing use case as a realistic, widely used scenario for document understanding, and propose a large scale DocVQA dataset comprising invoice documents and associated questions and answers. We employ a federated learning scheme, that reflects the real-life distribution of documents in different businesses, and we explore the use case where the ID of the invoice issuer is the sensitive information to be protected.
We demonstrate that non-private models tend to memorise, behaviour that can lead to exposing private information. We then evaluate baseline training schemes employing federated learning and differential privacy in this multi-modal scenario, where the sensitive information might be exposed through any of the two input modalities: vision (document image) or language (OCR tokens).
Finally, we design an attack exploiting the memorisation effect of the model, and demonstrate its effectiveness in probing different DocVQA models.

Abstract: Document Visual Question Answering (DocVQA) refers to the task of answering questions from document images. Existing work on DocVQA only considers single-page documents. However, in real scenarios documents are mostly composed of multiple pages that should be processed altogether. In this work we extend DocVQA to the multi-page scenario. For that, we first create a new dataset, MP-DocVQA, where questions are posed over multi-page documents instead of single pages. Second, we propose a new hierarchical method, Hi-VT5, based on the T5 architecture, that overcomes the limitations of current methods to process long multi-page documents. The proposed method is based on a hierarchical transformer architecture where the encoder summarizes the most relevant information of every page and then, the decoder takes this summarized information to generate the final answer. Through extensive experimentation, we demonstrate that our method is able, in a single stage, to answer the questions and provide the page that contains the relevant information to find the answer, which can be used as a kind of explainability measure.

Abstract: Visual Question Answering (VQA) is the task where given an image and a natural language question, the objective is to generate a natural language answer. At the intersection between computer vision and natural language processing, this task can be seen as a measure of image understanding capabilities, as it requires to reason about objects, actions, colors, positions, the relations between the different elements as well as commonsense reasoning, world knowledge, arithmetic skills and natural language understanding. However, even though the text present in the images conveys important semantically rich information that is explicit and not available in any other form, most VQA methods remained illiterate, largely
ignoring the text despite its potential significance. In this thesis, we set out on a journey to bring reading capabilities to computer vision models applied to the VQA task, creating new datasets and methods that can read, reason and integrate the text with other visual cues in natural scene images and documents.
In Chapter 3, we address the combination of scene text with visual information to fully understand all the nuances of natural scene images. To achieve this objective, we define a new sub-task of VQA that requires reading the text in the image, and highlight the limitations of the current methods. In addition, we propose a new architecture that integrates both modalities and jointly reasons about textual and visual features. In Chapter 5, we shift the domain of VQA with reading capabilities and apply it on scanned industry document images, providing a high-level end-purpose perspective to Document Understanding, which has been
primarily focused on digitizing the document’s contents and extracting key values without considering the ultimate purpose of the extracted information. For this, we create a dataset which requires methods to reason about the unique and challenging elements of documents, such as text, images, tables, graphs and complex layouts, to provide accurate answers in natural language. However, we observed that explicit visual features provide a slight contribution in the overall performance, since the main information is usually conveyed within the text and its position. In consequence, in Chapter 6, we propose VQA on infographic images, seeking for document images with more visually rich elements that require to fully exploit visual information in order to answer the questions. We show the performance gap of
different methods when used over industry scanned and infographic images, and propose a new method that integrates the visual features in early stages, which allows the transformer architecture to exploit the visual features during the self-attention operation. Instead, in Chapter 7, we apply VQA on a big collection of single-page documents, where the methods must find which documents are relevant to answer the question, and provide the answer itself. Finally, in Chapter 8, mimicking real-world application problems where systems must process documents with multiple pages, we address the multipage document visual question answering task. We demonstrate the limitations of existing methods, including models specifically designed to process long sequences. To overcome these limitations, we propose
a hierarchical architecture that can process long documents, answer questions, and provide the index of the page where the information to answer the question is located as an explainability measure.

Abstract: We propose a novel way to improve the generalisation capacity of deep learning models by reducing high correlations between neurons. For this, we present two regularisation terms computed from the weights of a minimum spanning tree of the clique whose vertices are the neurons of a given network (or a sample of those), where weights on edges are correlation dissimilarities. We provide an extensive set of experiments to validate the effectiveness of our terms, showing that they outperform popular ones. Also, we demonstrate that naive minimisation of all correlations between neurons obtains lower accuracies than our regularisation terms, suggesting that redundancies play a significant role in artificial neural networks, as evidenced by some studies in neuroscience for real networks. We include a proof of differentiability of our regularisers, thus developing the first effective topological persistence-based regularisation terms that consider the whole set of neurons and that can be applied to a feedforward architecture in any deep learning task such as classification, data generation, or regression.

Abstract: Confocal Laser Endomicroscopy (CLE) is an emerging imaging technique that allows the in-vivo acquisition of cell patterns of potentially malignant lesions. Such patterns could discriminate between inflammatory and neoplastic lesions and, thus, serve as a first in-vivo biopsy to discard cases that do not actually require a cell biopsy.
The goal of this work is to explore whether CLE images obtained during videobronchoscopy contain enough visual information to discriminate between benign and malign peripheral lesions for lung cancer diagnosis. To do so, we have performed a pilot comparative study with 12 patients (6 adenocarcinoma and 6 benign-inflammatory) using 2 different methods for CLE pattern analysis: visual analysis by 3 experts and a novel methodology that uses graph methods to find patterns in pre-trained feature spaces. Our preliminary results indicate that although visual analysis can only achieve a 60.2% of accuracy, the accuracy of the proposed unsupervised image pattern classification raises to 84.6%.
We conclude that CLE images visual information allow in-vivo detection of neoplastic lesions and graph structural analysis applied to deep-learning feature spaces can achieve competitive results.

Abstract: People are becoming increasingly concerned about their diet, whether for disease prevention, medical treatment or other purposes. In meals served in restaurants, schools or public canteens, it is not easy to identify the ingredients and/or the nutritional information they contain. Currently, technological solutions based on deep learning models have facilitated the recording and tracking of food consumed based on the recognition of the main dish present in an image. Considering that sometimes there may be multiple foods served on the same plate, food analysis should be treated as a multi-class object detection problem. EfficientDet and YOLOv5 are object detection algorithms that have demonstrated high mAP and real-time performance on general domain data. However, these models have not been evaluated and compared on public food datasets. Unlike general domain objects, foods have more challenging features inherent in their nature that increase the complexity of detection. In this work, we performed a performance evaluation of Efficient-Det and YOLOv5 on three public food datasets: UNIMIB2016, UECFood256 and ChileanFood64. From the results obtained, it can be seen that YOLOv5 provides a significant difference in terms of both mAP and response time compared to EfficientDet in all datasets. Furthermore, YOLOv5 outperforms the state-of-the-art on UECFood256, achieving an improvement of more than 4% in terms of mAP@.50 over the best reported.