As imaging technologies, both software and hardware, advance, they surpass traditional capabilities, capturing and interpreting visual information beyond the limits of human perception. While these cutting-edge computational imaging systems push the boundaries of what can be seen and understood, they also quietly introduce critical ethical concerns related to privacy, safety, and robustness. Since these systems operate beyond human vision, many potential threats remain imperceptible, making them more difficult to detect and mitigate.

Imaging Privacy Privacy concerns have gained increasing attention in the computer vision community, particularly as the progression of data-driven methods demands vast and often sensitive datasets. While traditional discussions have focused on visible privacy risks, advanced imaging systems can capture invisible yet private signals, introducing a new class of threats. Researchers have demonstrated how seemingly innocuous reflections/interreflections captured by a conventional camera—from a wall (Torralba et al, 2012, Sharma et al., 2021), a corner (Bouman et al., 2017), the human eyes (Nishino and Nayar, 2004, Alzayer et al., 2024), a bag of chips (Park et al., 2020), or a soda cans (Yu et al., 2023)—can be leveraged to reconstruct private scenes. Dynamic scenes encode more substantial information; for instance, we can extract hidden human motion (Aittala et al., 2019) and intention (Liu et al., 2024) as well as detect sound through vibrations in high-speed videos (Davis et al., 2014), enabling eavesdropping from a distance. Advanced cameras capable of capturing non-visible spectra (e.g., NIR, SWIR, TIR, Terahertz, Radio) further expand privacy threats. For example, the thermal spectra encoded the invisible human activity (Liu and Vondrick, 2023) and what happened before 3 seconds (Tang et al., 2023). The radio signal can be used to probe human activity through walls (Adib and Katabi, 2013, Zhao et al., 2018).

Imaging Safety Besides the privacy concerns, active imaging methods that emit energy, such as laser (O’Toole et al., 2018), UV (Krishnan and Fergus, 2009), radio waves (Adib and Katabi, 2013, Zhao et al., 2018, Lai et al., 2024), or sound waves (Xu and Wang, 2006), to illuminate the targets introduce security and safety risks. For example, systems that use high-energy lasers for depth sensing or scene reconstruction could inadvertently cause harm, posing safety threats to humans (Tilmon et al., 2023).

Imaging Robustness Ensuring the robustness of AI systems equipped with advanced imaging devices—capable of perceiving beyond human vision—poses significant challenges. For instance, it is possible to design attack patterns that are imperceptible to the human eye yet can effectively compromise these systems (Wei et al., 2024, Wei et al., 2023, Lin et al., 2025). Besides, the emitted energy from active imaging system can bring interference and lead to potential risks.

This workshop aims to bring attention to these challenges and explore innovative solutions to deal with these challengings as imaging technologies push the boundaries of perceiving the invisible.

• Novel Work: This track welcomes early-stage research that has not yet undergone peer review. We actively encourage the submission of works-in-progress and emerging ideas to foster discussion and collaboration among participants. Please use the ICCV 2025 template. Submissions in this track will be reviewed for relevance and soundness.
  • Proceedings: If you would like your work to be included in the ICCV proceedings, please submit through this OpenReview by July 3, 2025. Submissions should be between 4 and 8 pages in length, excluding references.
  • Non-Proceedings: If you prefer not to be included in the ICCV proceedings, please submit through this Openreview by August 19, 2025. Submissions may take one of the following forms (excluding references):
    • Full Paper: 4–8 pages
    • Short Paper: Up to 4 pages
    • Tiny Paper / Extended Abstract: Up to 2 pages
• Published Work: This track invites submissions of works that have already been accepted at previous venues (e.g., CVPR, ICCV, ECCV, IJCV, TPAMI, ICCP, TCI, SIGGRAPH, etc.), including those accepted to the ICCV 2025 main conference. Submissions can be in any format and should be sent directly via email: responsible.imaging@gmail.com