Abstract: Vision and Language are broadly regarded as cornerstones of intelligence. Even though language and vision have different aims – language having the purpose of communication, transmission of information and vision having the purpose of constructing mental representations around us to navigate and interact with objects – they cooperate and depend on one another in many tasks we perform effortlessly. This reliance is actively being studied in various Computer Vision tasks, e.g. image captioning, visual question answering, image-sentence retrieval, phrase grounding, just to name a few. All of these tasks share the inherent difficulty of the aligning the two modalities, while being robust to language
priors and various biases existing in the datasets. One of the ultimate goal for vision and language research is to be able to inject world knowledge while getting rid of the biases that come with the datasets. In this thesis, we mainly focus on two vision and language tasks, namely Image Captioning and Scene-Text Visual Question Answering (STVQA).
In both domains, we start by defining a new task that requires the utilization of world knowledge and in both tasks, we find that the models commonly employed are prone to biases that exist in the data. Concretely, we introduce new tasks and discover several problems that impede performance at each level and provide remedies or possible solutions in each chapter: i) We define a new task to move beyond Image Captioning to Image Interpretation that can utilize Named Entities in the form of world knowledge. ii) We study the object hallucination problem in classic Image Captioning systems and develop an architecture-agnostic solution. iii) We define a sub-task of Visual Question Answering that requires reading the text in the image (STVQA), where we highlight the limitations of current models. iv) We propose an architecture for the STVQA task that can point to the answer in the image and show how to combine it with classic VQA models. v) We show how far language can get us in STVQA and discover yet another bias which causes the models to disregard the image while doing Visual Question Answering.

Abstract: Multimodal learning from document data has achieved great success lately as it allows to pre-train semantically meaningful features as a prior into a learnable downstream approach. In this paper, we approach the document classification problem by learning cross-modal representations through language and vision cues, considering intra- and inter-modality relationships. Instead of merging features from different modalities into a common representation space, the proposed method exploits high-level interactions and learns relevant semantic information from effective attention flows within and across modalities. The proposed learning objective is devised between intra- and inter-modality alignment tasks, where the similarity distribution per task is computed by contracting positive sample pairs while simultaneously contrasting negative ones in the common feature representation space}. Extensive experiments on public document classification datasets demonstrate the effectiveness and the generalization capacity of our model on both low-scale and large-scale datasets.

Abstract: Nowadays the paradigm of link discovery problems has shown significant improvements on Knowledge Graphs. However, method performances are harmed by the unbalanced nature of this classification problem, since many methods are easily biased to not find proper links. In this paper we present a discriminative neural variational auto-encoder model, called DNVAE from now on, in which we have introduced latent variables to serve as embedding vectors. As a result, the learnt generative model approximate better the underlying distribution and, at the same time, it better differentiate the type of relations in the knowledge graph. We have evaluated this approach on benchmark knowledge graph and Census records. Results in this last data set are quite impressive since we reach the highest possible score in the evaluation metrics. However, further experiments are still needed to deeper evaluate the performance of the method in more challenging tasks.

Abstract: In this paper, we propose a new feature vector, named DElaunay TRIangulation-based Features (DETRIF), for real-time camera-based document image retrieval. DETRIF is computed based on the geometrical constraints from each pair of adjacency triangles in delaunay triangulation which is constructed from centroids of connected components. Besides, we employ a hashing-based indexing system in order to evaluate the performance of DETRIF and to compare it with other systems such as LLAH and SRIF. The experimentation is carried out on two datasets comprising of 400 heterogeneous-content complex linguistic map images (huge size, 9800 X 11768 pixels resolution)and 700 textual document images.

Abstract: Optical Music Recognition refers to the task of transcribing the image of a music score into a machine-readable format. Many music scores are written in a single staff, and therefore, they could be treated as a sequence. Therefore, this work explores the use of Long Short-Term Memory (LSTM) Recurrent Neural Networks for reading the music score sequentially, where the LSTM helps in keeping the context. For training, we have used a synthetic dataset of more than 40000 images, labeled at primitive level. The experimental results are promising, showing the benefits of our approach.