Abstract: Many scene text understanding methods approach the endtoend recognition problem from a word-spotting perspective and take huge benet from using small per-image lexicons. Such customized lexicons are normally assumed as given and their source is rarely discussed.
In this paper we propose a method that generates contextualized lexicons
for scene images using only visual information. For this, we exploit
the correlation between visual and textual information in a dataset consisting
of images and textual content associated with them. Using the topic modeling framework to discover a set of latent topics in such a dataset allows us to re-rank a xed dictionary in a way that prioritizes the words that are more likely to appear in a given image. Moreover, we train a CNN that is able to reproduce those word rankings but using only the image raw pixels as input. We demonstrate that the quality of the automatically obtained custom lexicons is superior to a generic frequency-based baseline.

Abstract: Many natural image sets, depicting objects whose appearance is changing due to motion, pose or light variations, can be considered samples of a low-dimension nonlinear manifold embedded in the high-dimensional observation space (the space of all possible images). The main contribution of our work is represented by a Supervised Laplacian Eigemaps (S-LE) algorithm, which exploits the class label information for mapping the original data in the embedded space. Our proposed approach benefits from two important properties: i) it is discriminative, and ii) it adaptively selects the neighbors of a sample without using any predefined neighborhood size. Experiments were conducted on four face databases and the results demonstrate that the proposed algorithm significantly outperforms many linear and non-linear embedding techniques. Although we've focused on the face recognition problem, the proposed approach could also be extended to other category of objects characterized by large variance in their appearance.

Abstract: Many historians and linguists are working individually and in an uncoordinated fashion on the identification and decryption of historical ciphers. This is a time-consuming process as they often work without access to automatic methods and processes that can accelerate the decipherment. At the same time, computer scientists and cryptologists are developing algorithms to decrypt various cipher types without having access to a large number of original ciphertexts. In this paper, we describe the DECRYPT project aiming at the creation of resources and tools for historical cryptology by bringing the expertise of various disciplines together for collecting data, exchanging methods for faster progress to transcribe, decrypt and contextualize historical encrypted manuscripts. We present our goals and work-in progress of a general approach for analyzing historical encrypted manuscripts using standardized methods and a new set of state-of-the-art tools. We release the data and tools as open-source hoping that all mentioned disciplines would benefit and contribute to the research infrastructure of historical cryptology.

Abstract: Manually annotating images to develop vision models has been a major bottleneck
since computer vision and machine learning started to walk together. This has
been more evident since computer vision falls on the shoulders of data-hungry
deep learning techniques. When addressing on-board perception for autonomous
driving, the curse of data annotation is exacerbated due to the use of additional
sensors such as LiDAR. Therefore, any approach aiming at reducing such a timeconsuming and costly work is of high interest for addressing autonomous driving
and, in fact, for any application requiring some sort of artificial perception. In the
last decade, it has been shown that leveraging from synthetic data is a paradigm
worth to pursue in order to minimizing manual data annotation. The reason is
that the automatic process of generating synthetic data can also produce different
types of associated annotations (e.g. object bounding boxes for synthetic images
and LiDAR pointclouds, pixel/point-wise semantic information, etc.). Directly
using synthetic data for training deep perception models may not be the definitive
solution in all circumstances since it can appear a synth-to-real domain shift. In
this context, this work focuses on leveraging synthetic data to alleviate manual
annotation for three perception tasks related to driving assistance and autonomous
driving. In all cases, we assume the use of deep convolutional neural networks
(CNNs) to develop our perception models.
The first task addresses traffic sign recognition (TSR), a kind of multi-class
classification problem. We assume that the number of sign classes to be recognized
must be suddenly increased without having annotated samples to perform the
corresponding TSR CNN re-training. We show that leveraging synthetic samples of
such new classes and transforming them by a generative adversarial network (GAN)
trained on the known classes (i.e. without using samples from the new classes), it is
possible to re-train the TSR CNN to properly classify all the signs for a ∼ 1/4 ratio of
new/known sign classes. The second task addresses on-board 2D object detection,
focusing on vehicles and pedestrians. In this case, we assume that we receive a set
of images without the annotations required to train an object detector, i.e. without
object bounding boxes. Therefore, our goal is to self-annotate these images so
that they can later be used to train the desired object detector. In order to reach
this goal, we leverage from synthetic data and propose a semi-supervised learning
approach based on the co-training idea. In fact, we use a GAN to reduce the synthto-real domain shift before applying co-training. Our quantitative results show
that co-training and GAN-based image-to-image translation complement each
other up to allow the training of object detectors without manual annotation, and still almost reaching the upper-bound performances of the detectors trained from
human annotations. While in previous tasks we focus on vision-based perception,
the third task we address focuses on LiDAR pointclouds. Our initial goal was to
develop a 3D object detector trained on synthetic LiDAR-style pointclouds. While
for images we may expect synth/real-to-real domain shift due to differences in
their appearance (e.g. when source and target images come from different camera
sensors), we did not expect so for LiDAR pointclouds since these active sensors
factor out appearance and provide sampled shapes. However, in practice, we have
seen that it can be domain shift even among real-world LiDAR pointclouds. Factors
such as the sampling parameters of the LiDARs, the sensor suite configuration onboard the ego-vehicle, and the human annotation of 3D bounding boxes, do induce
a domain shift. We show it through comprehensive experiments with different
publicly available datasets and 3D detectors. This redirected our goal towards the
design of a GAN for pointcloud-to-pointcloud translation, a relatively unexplored
topic.
Finally, it is worth to mention that all the synthetic datasets used for these three
tasks, have been designed and generated in the context of this PhD work and will
be publicly released. Overall, we think this PhD presents several steps forward to
encourage leveraging synthetic data for developing deep perception models in the
field of driving assistance and autonomous driving.

Abstract: Manifold learning techniques are affected by two critical aspects: (i) the design of the adjacency graphs, and (ii) the embedding of new test data---the out-of-sample problem. For the first aspect, the proposed schemes were heuristically driven. For the second aspect, the difficulty resides in finding an accurate mapping that transfers unseen data samples into an existing manifold. Past works addressing these two aspects were heavily parametric in the sense that the optimal performance is only reached for a suitable parameter choice that should be known in advance. In this paper, we demonstrate that sparse coding theory not only serves for automatic graph reconstruction as shown in recent works, but also represents an accurate alternative for out-of-sample embedding. Considering for a case study the Laplacian Eigenmaps, we applied our method to the face recognition problem. To evaluate the effectiveness of the proposed out-of-sample embedding, experiments are conducted using the k-nearest neighbor (KNN) and Kernel Support Vector Machines (KSVM) classifiers on four public face databases. The experimental results show that the proposed model is able to achieve high categorization effectiveness as well as high consistency with non-linear embeddings/manifolds obtained in batch modes.

Abstract: Major advances have recently been made in merging language and vision representations. Most tasks considered so far have confined themselves to the processing of objects and lexicalised relations amongst objects (content words). We know, however, that humans (even pre-school children) can abstract over raw multimodal data to perform certain types of higher level reasoning, expressed in natural language by function words. A case in point is given by their ability to learn quantifiers, i.e. expressions like few, some and all. From formal semantics and cognitive linguistics, we know that quantifiers are relations over sets which, as a simplification, we can see as proportions. For instance, in most fish are red, most encodes the proportion of fish which are red fish. In this paper, we study how well current neural network strategies model such relations. We propose a task where, given an image and a query expressed by an object–property pair, the system must return a quantifier expressing which proportions of the queried object have the queried property. Our contributions are twofold. First, we show that the best performance on this task involves coupling state-of-the-art attention mechanisms with a network architecture mirroring the logical structure assigned to quantifiers by classic linguistic formalisation. Second, we introduce a new balanced dataset of image scenarios associated with quantification queries, which we hope will foster further research in this area.

Abstract: Machine learning experimentation under controlled scenarios and standard datasets is necessary to compare algorithms performance by evaluating all of them in the same setup. However, experimentation on how those algorithms perform on unconstrained data and applied tasks to solve real world problems is also a must to ascertain how that research can contribute to our society.
In this dissertation we experiment with the latest computer vision and natural language processing algorithms applying them to multimodal scene interpretation. Particularly, we research on how image and text understanding can be jointly exploited to address real world problems, focusing on learning from Social Media data.
We address several tasks that involve image and textual information, discuss their characteristics and offer our experimentation conclusions. First, we work on detection of scene text in images. Then, we work with Social Media posts, exploiting the captions associated to images as supervision to learn visual features, which we apply to multimodal semantic image retrieval. Subsequently, we work with geolocated Social Media images with associated tags, experimenting on how to use the tags as supervision, on location sensitive image retrieval and on exploiting location information for image tagging. Finally, we work on a specific classification problem of Social Media publications consisting on an image and a text: Multimodal hate speech classification.

Abstract: Low resource Handwritten Text Recognition (HTR) is a hard problem due to the scarce annotated data and the very limited linguistic information (dictionaries and language models). This appears, for example, in the case of historical ciphered manuscripts, which are usually written with invented alphabets to hide the content. Thus, in this paper we address this problem through a data generation technique based on Bayesian Program Learning (BPL). Contrary to traditional generation approaches, which require a huge amount of annotated images, our method is able to generate human-like handwriting using only one sample of each symbol from the desired alphabet. After generating symbols, we create synthetic lines to train state-of-the-art HTR architectures in a segmentation free fashion. Quantitative and qualitative analyses were carried out and confirm the effectiveness of the proposed method, achieving competitive results compared to the usage of real annotated data.