Welcome!
I'm a grad student at the National University of Colombia. My research area is in the field of Information Retrieval, specially, image retrieval. My main research interests are image retrieval, machine learning, computer vision and distributed computing.
Research Work
Content-Based Image Retrieval
Finding images in large collections may be a very time-consuming and frustrating task. In this work, we investigated the potential of image search tools based on visual features rather than associated text. Users are expected to provide an example image and the system will retrieve the most similar ones from the collection. [More]
Automatic Image Annotation
Large collections of images do not have a detailed description of what people can see in every picture. Some images come with certain annotations and some others do not even have any. In this research, the problem of assigning semantic labels to images by analysing visual contents is approached. These labels can be used in a keyword-based image search engine. [More]
Feature Combination and Fusion
Be it for content-based image retrieval or image classification, visual contents need to be represented using expressive enough features. There are different kinds of features, each describing some visual properties in images. This research aims to build models to combine many visual features in an integrated image representation to improve retrieval and classification performance. [More]
Multimodal Indexing of Image Collections
Modern image collections have plenty of associated metadata, such as GPS geo-tagging, user's comments, ratings, text descriptions and so on. Since the semantic gap has been found to be one of the main problems in image retrieval based on visual contents only, other data modalities can be used to complement and to improve search results. We research the problem of retrieving images by combining visual features and associated text descriptions in the same indexing model. [More]
Image Enhancement
Taking the perfect picture is a challenging task for average users using general purpose cameras. Yet, retouching every single photo in an album is a very time consuming task. That's why users usually do not enhance personal images, but rather archive them even if they're not in their best quaility. In this project we proposed a system that recommeds enhancements in a collaborative environment, i.e., looking for users with similar enhancement preferences, to infeer the parameters required to produce a retouched image.
