Bio
I am an Assistant Professor at the Autonomous University of Barcelona and a researcher at the Computer Vision Center in Barcelona, specialized in representation learning. My research focuses on the emerging properties in foundation models.
I have published work in top venues, including CVPR, ECCV, NeurIPS, and ICLR. I completed my PhD at the Autonomous University of Barcelona under the supervision of Prof. Joost van de Weijer. I received my Master of Science degree in 2018 from SUPSI, Switzerland, and my Bachelor's degree in Electronics Engineering in 2016 from Universidad de Antioquia, Colombia.
Research interests
- Emerging properties in foundation models
- Representation learning
- Human and machine visual perception
- Generative models for creative applications
News
- NeurIPSTwo papers accepted at NeurIPS 2026: DiDE on color-texture decoupling for 3D stylization, and Disentangling the Good From the Bad on quantization-induced flips.
- ECCVTwo papers accepted at ECCV 2026: DriftScope and NumColor.
- CICOur paper Do Vision Encoders Exhibit Human-like Color Thresholds? was accepted at the Color and Imaging Conference 2026.
- JournalOur paper Contrast sensitivity in multimodal large language models: A psychophysics-inspired evaluation was published in Neural Networks.
- ICMLTwo papers at ICML 2026: Can Less Precise Be More Reliable? on quantization and CLIP, and Online Continual Learning with Dynamic Label Hierarchies.
- CVPROur paper GenColorBench: A Color Evaluation Benchmark for Text-to-Image Generation was presented at CVPR 2026.
- WACVOur paper Leveraging Semantic Attribute Binding for Free-Lunch Color Control in Diffusion Models was accepted at WACV 2026.
- NeurIPSOur paper Free-Lunch Color-Texture Disentanglement for Stylized Image Generation was accepted at NeurIPS 2025.
- PressOur CVPR paper was discussed in a Communications of the ACM article.
- GrantWe received a EuroHPC development grant for our open-world forgetting project.
- UABJoining UAB/CVC as an Assistant Professor in Fall 2025.
- CoLLAsOur paper CLA: Latent Alignment for Online Continual Self-Supervised Learning was accepted at CoLLAs 2025.
- CVPROur paper The Art of Deception: Color Visual Illusions and Diffusion Models was accepted at CVPR 2025.
- PhDI defended my PhD thesis, The Analysis and Continual Learning of Deep Feature Representations, with the Cum Laude distinction.
- ECCVOur paper Exemplar-free continual representation learning via learnable drift compensation was accepted at ECCV 2024.
- WACVOur paper Plasticity-optimized complementary networks for unsupervised continual learning was accepted at WACV 2024.
- ICLROur paper Planckian jitter: countering the color-crippling effects of color jitter on self-supervised training was accepted at ICLR 2023.
- AwardOur paper Continually learning self-supervised representations with projected functional regularization received the best paper award at the CVPR 2022 continual learning workshop.
Publications
Conferences 20
- NeurIPS2026DiDE: Direct Injection with Color-Texture DEcoupling for 3D StylizationAdvances in Neural Information Processing Systems
- NeurIPS2026Disentangling the Good From the Bad: Quantization-Induced Flips Are Not RandomAdvances in Neural Information Processing Systems
- ECCV2026DriftScope: Measuring The Hidden Effects of Diffusion Model AdaptationEuropean Conference on Computer Vision
- ECCV2026NumColor: Precise Numeric Color Control in Text-to-Image GenerationEuropean Conference on Computer Vision
- ICML2026Can Less Precise Be More Reliable? A Systematic Evaluation of Quantization’s Impact on CLIP Beyond AccuracyInternational Conference on Machine Learning
- ICML2026Online Continual Learning with Dynamic Label HierarchiesInternational Conference on Machine Learning
- CVPR2026GenColorBench: A Color Evaluation Benchmark for Text-to-Image GenerationIEEE/CVF Conference on Computer Vision and Pattern Recognition
- CIC2026Do Vision Encoders Exhibit Human-like Color Thresholds?Color and Imaging Conference
- WACV2026Leveraging Semantic Attribute Binding for Free-Lunch Color Control in Diffusion ModelsIEEE/CVF Winter Conference on Applications of Computer Vision
- NeurIPS2025Free-Lunch Color-Texture Disentanglement for Stylized Image GenerationAdvances in Neural Information Processing Systems
- CoLLAs2025CLA: Latent Alignment for Online Continual Self-Supervised LearningFourth Conference on Lifelong Learning Agents
- CVPR2025The Art of Deception: Color Visual Illusions and Diffusion ModelsIEEE/CVF Conference on Computer Vision and Pattern Recognition
- ECCV2024Exemplar-free continual representation learning via learnable drift compensationEuropean Conference on Computer Vision, pp. 473–490
- WACV2024Plasticity-optimized complementary networks for unsupervised continual learningIEEE/CVF Winter Conference on Applications of Computer Vision, pp. 1690–1700
- ICLR2023Planckian jitter: countering the color-crippling effects of color jitter on self-supervised trainingInternational Conference on Learning Representations
- CVPR-W2022Continually learning self-supervised representations with projected functional regularizationCVPR Workshops, pp. 3867–3877 · Best paper award
- CVPR2019Convolutional neural networks can be deceived by visual illusionsIEEE/CVF Conference on Computer Vision and Pattern Recognition, pp. 12309–12317
- ICTAI2017Recognition of mammal genera on camera-trap images using multi-layer robust principal component analysis and mixture neural networksIEEE International Conference on Tools with Artificial Intelligence, pp. 53–60
- ISVC2016Animal identification in low quality camera-trap images using very deep convolutional neural networks and confidence thresholdsInternational Symposium on Visual Computing, pp. 747–756
- ISVC2015A Markov random field and active contour image segmentation model for animal spots patternsInternational Symposium on Visual Computing, pp. 173–184
Journals 10
- Neural Netw.2026
- J. Vision2022On the synthesis of visual illusions using deep generative modelsJournal of Vision, 22(8)
- J. Vision2022Contrast sensitivity functions in autoencodersJournal of Vision, 22(6)
- Vis. Res.2020Color illusions also deceive CNNs for low-level vision tasks: Analysis and implicationsVision Research, 176, 156–174
- Sci. Rep.2020Evidence for the intrinsically nonlinear nature of receptive fields in visionScientific Reports, 10(1), 16277
- J. Neuro.2020Visual information flow in Wilson–Cowan networksJournal of Neurophysiology, 123(6), 2249–2268
- J. Neuro.2020Visual illusions via neural dynamics: Wilson–Cowan-type models and the efficient representation principleJournal of Neurophysiology, 123(5), 1606–1618
- Vis. Comp.2019Camera-trap images segmentation using multi-layer robust principal component analysisThe Visual Computer, 35, 335–347
- PAA2018Automatic identification of Scenedesmus polymorphic microalgae from microscopic imagesPattern Analysis and Applications, 21, 601–612
- Eco. Inf.2017Towards automatic wild animal monitoring: Identification of animal species in camera-trap images using very deep convolutional neural networksEcological Informatics, 41, 24–32
Workshops and others 2
- arXiv2024At the edge of a generative cultural precipicearXiv:2406.08739
- arXiv2018
Under review 3
- Preprint2026On Color Alignment in VAE Latent Spaces and Its ApplicationsarXiv:2610.07072
- Preprint2026
- Preprint2026
Teaching
- Autumn 2025Data Structures
- Spring 20253D Vision and Motion Analysis