Raesetje Sefala

Jabez Magomere, Shu Ishida, Tejumade Afonja et al.

We introduce the World Wide recipe, which sets forth a framework for culturally aware and participatory data collection, and the resultant regionally diverse World Wide Dishes evaluation dataset. We also analyse bias operationalisation to highlight how current systems underperform across several dimensions: (in-)accuracy, (mis-)representation, and cultural (in-)sensitivity, with evidence from qualitative community-based observations and quantitative automated tools. We find that these T2I models generally do not produce quality outputs of dishes specific to various regions. This is true even for the US, which is typically considered more well-resourced in training data -- although the generation of US dishes does outperform that of the investigated African countries. The models demonstrate the propensity to produce inaccurate and culturally misrepresentative, flattening, and insensitive outputs. These representational biases have the potential to further reinforce stereotypes and disproportionately contribute to erasure based on region. The dataset and code are available at https://github.com/oxai/world-wide-dishes.

João Caldeira, Alex Fout, Aniket Kesari et al.

This project presents the results of a partnership between the Data Science for Social Good fellowship, Jakarta Smart City and Pulse Lab Jakarta to create a video analysis pipeline for the purpose of improving traffic safety in Jakarta. The pipeline transforms raw traffic video footage into databases that are ready to be used for traffic analysis. By analyzing these patterns, the city of Jakarta will better understand how human behavior and built infrastructure contribute to traffic challenges and safety risks. The results of this work should also be broadly applicable to smart city initiatives around the globe as they improve urban planning and sustainability through data science approaches.

Michael Burke, Siyabonga Mbonambi, Purity Molala et al.

A great deal of work aims to discover large general purpose models of image interest or memorability for visual search and information retrieval. This paper argues that image interest is often domain and user specific, and that efficient mechanisms for learning about this domain-specific image interest as quickly as possible, while limiting the amount of data-labelling required, are often more useful to end-users. This work uses pairwise image comparisons to reduce the labelling burden on these users, and introduces an image interest estimation approach that performs similarly to recent data hungry deep learning approaches trained using pairwise ranking losses. Here, we use a Gaussian process model to interpolate image interest inferred using a Bayesian ranking approach over image features extracted using a pre-trained convolutional neural network. Results show that fitting a Gaussian process in high-dimensional image feature space is not only computationally feasible, but also effective across a broad range of domains. The proposed probabilistic interest estimation approach produces image interests paired with uncertainties that can be used to identify images for which additional labelling is required and measure inference convergence, allowing for sample efficient active model training. Importantly, the probabilistic formulation allows for effective visual search and information retrieval when limited labelling data is available.