Dear General Vision Sciences & Engineering communities,
The goal of the Fourth Workshop on Shared Visual Representations in Human and Machine Intelligence (SVRHM) at NeurIPS 2022 is to discuss and disseminate relevant findings and parallels between the computational neuro/cognitive science and machine learning/artificial intelligence communities.
In the past few years, machine learning tools — especially deep neural networks — have permeated the vision/cognitive/neuroscience communities to become the leading computational models that describe many cognitive tasks. Huge strides are also being made in the machine learning/artificial intelligence community with biologically inspired algorithms providing large efficiency gains in both computational and learning capabilities. However, many mysteries remain with regards to the alignment of human and machine perception, and there are cases where we see divergent rather than convergent representations. To resolve such questions, this workshop aims to bring fruitful discussions between scientists and engineers with multi-disciplinary backgrounds to review the recent progress in shared visual representations in both humans and machines, and in doing so identifying road-blocks and areas of interest to further accelerate the growth of both fields.
The workshop will include a series of talks and panel discussions from a diverse group of speakers from both industry and academia who will share their research at the intersection of humans and machines that pushes the field of vision forward. The aim of our Call for Papers is to bring together scientists and engineers to share their work in progress at the Poster Session that are applicable to the scope of the Workshop. This year in SVRHM, the 6 highest scoring papers will also be awarded an Oral Presentation entry in the program.
The following areas provide a sense of suitable topics for 4 to 5 page paper submissions:
- Biological inspiration and inductive bias in vision
- Human-relevant strategies for robustness and generalization
- New datasets (e.g., for comparing humans/animals and machines)
- Biologically-driven self-supervision
- Perceptual invariance and metamerism
- Biologically-informed strategies to mitigate adversarial vulnerability
- Foveation, active perception, and attention models
- Intuitive physics
- Biologically inspired Generative Models
- Perceptual and cognitive robustness
- Nuances and noise in perceptual and cognitive systems
- Creative problem-solving
- Differences and similarities between humans and deep neural networks
- Canonical computations in biological and artificial systems
- Alternative architectures for deep neural networks
- Reverse engineering of the human visual system via deep neural networks
- New: Biologically-inspired Applications to Augmented and/or Virtual Reality
- New: Understanding of novel hand-engineered models e.g. Transformers and Capsule Networks
- New: Computational Aesthetics/Memorability/Humor/Virality
We will be awarding a new NVIDIA Titan GPU as the AI Best Paper Award, and depending on the levels of attendance will also provide reimbursements for first authors of accepted papers (to be confirmed), in addition to a tentative evening reception.
Link to the workshop with additional details for the Call for Papers: https://www.svrhm.com / https://www.svrhm.com/call-for-papers
Link to Paper workshop submission: https://openreview.net/group?id=NeurIPS.cc/2022/Workshop/SVRHM
The submission deadline is September 24th, 11.59pm PST, 2022.
Paper Acceptance: There will be no cap on paper acceptance percentage and our criteria for acceptance will continue to be to accept all papers with an average (1-10) score of 5.5 and above. This strict cut-off removes many conflicts of interest in addition to further discussion windows that due to the time horizon are not possible. Note: Paper And Reviewing standards are still very high and there is currently an approximate 60% acceptance rate with very high quality works that are later published in Tier-1 venues.
Reviewing: If you are interested in becoming a Reviewer and have published/reviewed for this or similar venues before, please feel free to contact us as we are aiming for providing 3-4 reviews per submitted paper, where each reviewer will review no more than 2 papers.
We are also continually looking for more sponsors that may provide Best Paper or Best Reviewer Awards. If you are interested in getting involved, let us know.
All questions & inquiries regarding the workshop should be sent to: email@example.com , our Twitter account: @svrhm_workshop (preferred) and/or by directly contacting any of the organizers.
The workshop will happen *in-person* this year on December 2nd 2022 in New Orleans, Louisiana; USA. Virtual Attendance options will also be provided.
Looking forward to seeing you all!
Arturo Deza, Joshua Peterson, Apruva Ratan Murty, Tom Griffiths
The SVRHM workshop is currently sponsored by MIT’s Center from Brains, Minds and Machines (CBMM), the National Science Foundation (NSF), Artificio & NVIDIA.