Description
- Abstract:
- Discrepancies between human and deep neural network (DNN) strategies for object recognition pose critical challenges in domains demanding interpretable and trustworthy models. To address this gap, we introduce ClickMe v2, a large-scale human-in-the-loop data collection paradigm that captures category-diagnostic visual features through a gamified interface. In contrast to the original ClickMe, our new pipeline scales to the full ImageNet Large Scale Visual Recognition Challenge (ILSVRC) 2012 dataset, aggregates feature importance maps from an additional order of magnitude of global participants, and enforces rigorous data quality controls via catch trials and automated bot detection. This expansion enables a comprehensive examination of how humans prioritize visual information across diverse object categories. By making these large-scale human-derived saliency maps publicly available, ClickMe v2 provides a foundation for benchmarking model interpretability and guiding the development of biologically inspired vision systems.
- Notes:
- Senior thesis (ScB)--Brown University, 2025
- Concentration: Computational Neuroscience
Access Conditions
- Use and Reproduction
- All rights reserved
- Rights
- In Copyright
- Restrictions on Use
- All Rights Reserved
Citation
Gopal, Jay,
"ClickMe: Large-Scale Human-in-the-Loop Feature Importance Mapping to Train Brain-Aligned Deep Neural Networks"
(2025).
Cognitive, Linguistic, and Psychological Sciences Theses and Dissertations.
Brown Digital Repository. Brown University Library.
https://doi.org/10.26300/6a0j-ay62
Relations
Collection:
-
Cognitive, Linguistic, and Psychological Sciences Theses and Dissertations
Theses and Dissertations for the Cognitive, Linguistic, and Psychological Sciences department....