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: We propose a method to acquire multispectral data with handheld devices with front-mounted RGB cameras. We propose to use the display of the device as an illuminant while the camera captures images illuminated by the red, green and
blue primaries of the display. Three illuminants and three response functions of the camera lead to nine response values which are used for reflectance estimation. Results are promising and show that the accuracy of the spectral reconstruction improves in the range from 30-40% over the spectral
reconstruction based on a single illuminant. Furthermore, we propose to compute sensor-illuminant aware linear basis by discarding the part of the reflectances that falls in the sensorilluminant null-space. We show experimentally that optimizing reflectance estimation on these new basis functions decreases
the RMSE significantly over basis functions that are independent to sensor-illuminant. We conclude that, multispectral data acquisition is potentially possible with consumer hand-held devices such as tablets, mobiles, and laptops, opening up applications which are currently considered to be unrealistic.

Abstract: Nowadays, document forgery is becoming a real issue. A large amount of documents that contain critical information as payment slips, invoices or contracts, are constantly subject to fraudster manipulation because of the lack of security regarding this kind of document. Previously, a system to detect fraudulent documents based on its intrinsic features has been presented. It was especially designed to retrieve copy-move forgery and imperfection due to fraudster manipulation. However, when a set of characters is not present in the original document, copy-move forgery is not feasible. Hence, the fraudster will use a text toolbox to add or modify information in the document by imitating the font or he will cut and paste characters from another document where the font properties are similar. This often results in font type errors. Thus, a clue to detect document forgery consists of finding characters, words or sentences in a document with font properties different from their surroundings. To this end, we present in this paper an automatic forgery detection method based on document font features. Using the Conditional Random Field a measurement of probability that a character belongs to a specific font is made by comparing the character font features to a knowledge database. Then, the character is classified as a genuine or a fake one by comparing its probability to belong to a certain font type with those of the neighboring characters.