research

A 6 months end of studies internship on the problem of Short Term Action Anticipation on egocentric videos from the Ego4D dataset. Use of a Transformer in place of a ROI Pooling operation earning a 3rd place on the public leaderboard of the associated challenge.

A 3 months group project consisting of an otherview on robustness and explainability of Machine Learning methods with an accent on adversarial attacks and defenses.

A collaboration with the team developing DeepHyper on an 11 months contract as a research aide. Development of benchmarks and analytical tools to help improving the package followed by an application of its AutoML methods to a real-case Deep Learning problem.

2 months internship on the implementation of YOLOv2 and the test on this model of a semi-supervised learning method for object classification.