Audo-Sight: Enabling Ambient Interaction For Blind And Visually Impaired Individuals

Bhanuja Ainary

REVIEW 4 major objections 5 minor 1 cited by

Audo-Sight: Enabling Ambient Interaction For Blind And Visually Impaired Individuals

T0 review · 4 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Audo-Sight claims that a single voice-first, multimodal-large-language-model system can give blind and visually impaired users real-time, context-aware spoken assistance in both private and public settings, with age-based filtering and…

desk verdict A systems thesis with one good routing idea and a public-mode safety feature that is not implemented as claimed—worth a serious referee only if heavy revision is expected. read the letter →

arxiv 2505.00153 v1 pith:GIQF7BZY submitted 2025-04-30 cs.HC cs.DC

Bhanuja Ainary This is my paper · ORCID

classification cs.HCcs.DC

keywords assistivetechnologyblindandvisuallyimpairedmultimodallargelanguagemodelsvoiceinteractionlatency-awareroutingage-basedcontentfilteringblind-friendlyresponsegenerationSmartSightintegration

verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

Audo-Sight is a proposed assistive system that lets blind and visually impaired users ask spoken questions about their surroundings and receive spoken answers generated by a multimodal large language model. The thesis claims the system works in two modes: a private mode that identifies the registered user and adapts to their preferences, and a public mode for museums, malls, or kiosks where a shared installation serves whoever speaks to it. In public mode the system estimates whether the speaker is under 18 and passes queries through a safety filter, and it rewrites model output to avoid visual language such as colors, vague directions, and ableist phrasing. The payoff claimed is that assistive technology moves from passive object identification to interactive, conversational, context-aware guidance.

What carries the argument

The load-bearing components are the Cognition Engine, the Latency-Aware LLM Router, the BVI-Friendly Response Generator built on NeMo Guardrails, and, in public mode, the Age-Range Detector and Safe Query Filter. The Latency-Aware LLM Router measures whether the multimodal reasoning engine will respond in time; if it would be too slow, the system shortcuts to an Image-to-Text module feeding a text-only reasoner, and if latency is acceptable it lets the full multimodal path run. The BVI-Friendly Response Generator rewrites raw model output so that responses avoid phrases like 'as you can see,' avoid color-based identification unless functionally necessary, replace vague deictic expressions with concrete directions, and filter offensive or patronizing language. The public-mode Age-Range Detector uses a CNN-LSTM hybrid on voice features to classify users as under or over 18, and the Safe Query Filter, implemented with Llama Guard 3-1B, blocks unsafe queries before they reach the reasoning layer.

What would settle it

Run the Age-Range Detector on a held-out corpus of voice recordings labeled by true age around the 18-year boundary and compare its classifications to ground truth; if accuracy is near chance or systematically wrong for certain accents or audio conditions, the public mode's safety gate fails. A second check is to measure end-to-end active-mode latency on a low-end edge device with natural user queries: the claimed conversational real-time interaction would be falsified if median response time regularly exceeds a few seconds.

Watch

Extended reading notes

Core claim

The central claim is that an MLLM-driven, voice-first architecture can provide expedient, context-aware ambient interaction for blind and visually impaired individuals without manual reconfiguration. The system fuses voice input, image input from wearable glasses or fixed cameras, and text into a Fusion Hub, then a Cognition Engine chooses between a full multimodal reasoning path and a faster text-only path depending on measured latency. Output is passed through a BVI-Friendly Response Generator that removes visual references, replaces vague directions with concrete spatial descriptions, and blocks offensive terms. In public spaces, an Age-Range Detector classifies the speaker as under or over 18 and a Safe Query Filter blocks unsafe queries before they reach the reasoning engine. The thesis further claims that integrating this active conversational layer with SmartSight's passive object, face, and text recognition produces ambient guidance: the system alerts the user to environmental changes while also answering open-ended questions about what it sees.

Load-bearing premise

The public mode's safety guarantee depends on the CNN-LSTM age-range detector reliably classifying users as under or over 18 from voice alone, but the thesis reports no accuracy data, so a misclassification would either expose minors to inappropriate content or block adults from useful information.

Editorial extensions

If this is right

A blind or visually impaired user in a museum could ask open-ended follow-up questions about an exhibit and receive spoken answers grounded in the current visual scene, rather than one-shot captions.
A shared public installation could serve different users in sequence with no per-user calibration, using voice-triggered interaction, age estimation, and built-in content filtering.
The latency-aware router could keep interactions conversational on modest edge hardware by falling back to text-only reasoning only when the multimodal path is slow.
Blind-friendly response rewriting could measurably reduce inaccessible outputs such as color-dependent directions, vague 'over there' statements, and ableist phrasing.
Integrating active conversational assistance with passive object and face detection would let the system both proactively alert users to changes and answer specific queries, sharing one text-to-speech pipeline.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

The public mode's safety guarantee stands or falls on the age-from-voice classifier's accuracy, for which the thesis reports no numbers; that accuracy plausibly varies with accent, age near the 18-year boundary, background noise, and microphone quality, so a field evaluation across those conditions would be the natural next test.
The reported correlation between MLLM latency and the frequency of output vocabulary suggests a design possibility the author leaves implicit: routing could be tuned not only by query complexity but by expected answer vocabulary, or prompts could be constrained to keep outputs in common words for faster response.
The fixed-microphone public scenario assumes the environment's visual layout is already known or pre-mapped; extending the same interaction to genuinely open, unmodeled spaces would require pairing the public architecture with live visual input, an integration the thesis only partially explores.
The private mode's user-identification-by-voice design implies a testable tension: if the system refuses to respond to unknown voices, then any misidentification or enrollment error could lock a legitimate user out, so end-to-end authentication accuracy deserves the same scrutiny as the age classifier.

Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Desk Editor's Note

read the letter

I read the Audo-Sight thesis. Bottom line: the private-mode architecture has one genuinely useful idea—latency-aware routing between a multimodal LLM and a text-only LLM—and the public-mode safety feature has a load-bearing gap between the prose and the architecture diagram. With revision it could become a solid short systems paper; as is, it is not there yet.

What is new and what does well: the latency-aware router that switches between LLaVA and Llama 3.2 based on predicted response time is a sensible contribution and is described well enough to reimplement. The NeMo-Guardrails-based rewriting for BVI-friendly responses is a practical idea, and the concrete list of offensive terms and instructions in Section 3.2.2.7 shows real engagement with accessibility. The SmartSight prototype chapter is also refreshingly specific about hardware, networking, and passive/active mode behavior. The observation in Chapter 5 that MLLM latency seems to track the rarity of output vocabulary is an empirical lead worth following, though it is not yet backed by statistics.

Where it falls down: the evaluation is mostly latency numbers from a single M2 Mac with no baselines, error bars, or statistical tests. The bar chart in Figure 3.3 lacks any methodology—how many queries, which model, what prompt set? The Ngram correlation in Figure 5.7 is eyeballed, not measured. The bigger problem is the public mode. Section 4.3 promises that the Safe Query Filter 'alters the responses to match the provided age category,' but the architecture in Figure 4.1 only routes under-18 queries through a block filter; nothing downstream modifies the response to a simpler, age-appropriate version. The CNN-LSTM age detector itself has zero accuracy data, and voice-based age estimation is known to be unreliable near the 18-year boundary. So the central safety claim—age-appropriate content for children—is both unmeasured and not implemented as drawn. That is a load-bearing inconsistency, not a minor omission.

Who this is for: researchers working on edge-LLM systems for accessibility, or anyone building on SmartSight. The private-mode architecture and the latency routing could inform future work, but I would not trust the public-mode claims without a redesign and a user study.

Recommendation: send it to peer review only if the venue expects major revision; otherwise desk-reject on the strength of the evaluation. If the age-response gap is closed and the evaluation is redone with proper methodology, it could become a useful paper.

Referee Report

4 major / 5 minor

Summary. The manuscript presents Audo-Sight, an assistive system for blind and visually impaired (BVI) users that combines voice input, camera imagery, multimodal large language models (MLLMs), latency-aware routing, and NeMo-Guardrails-based response rewriting. A private mode provides user-specific interaction through voice identification, while a public mode replaces user identification with an age-range detector and a safe query filter. The thesis also describes an integration with the SmartSight wearable prototype and reports latency measurements from a single hardware setup (Section 3.3 and Section 5.4), a bar chart on the effect of the BVI-friendly response generator (Figure 3.3), and an exploratory Ngram-based correlation between MLLM latency and output-word frequency (Figure 5.7).

Significance. If the claims were fully supported, Audo-Sight would address a genuine gap: interactive, context-aware, voice-based assistance for BVI users in both private and public spaces, with a plausible safety mechanism for minors. The work has clear strengths: it builds on standard, reproducible components (Whisper, YOLO, Llama 3.2 Vision, Llama Guard, NeMo Guardrails), provides a concrete hardware/software stack in Table 3.1, uses the VizWiz-VQA dataset as a realistic evaluation input, and describes a working SmartSight prototype. However, the central claims of effectiveness and safety currently rest almost entirely on architectural description and informal latency measurements, not on controlled evaluation. The age-based safety mechanism is both unmeasured and, as described, not wired into the response-generation path. The significance is therefore conditional on substantial additional validation and on closing the architectural gap in the public-mode safety pipeline.

major comments (4)

[§4.4.1] The Age-Range Detector is a load-bearing component for the public-mode safety claim, but no accuracy data of any kind are provided. The section states that a CNN-LSTM model classifies users as under or over 18 from voice alone, yet it reports no training data, no validation set, no confusion matrix, and no evaluation. This matters directly: if a minor is misclassified as an adult, the query bypasses the Safe Query Filter entirely; if an adult is misclassified as a minor, the adult is unnecessarily blocked. Voice-based age estimation near the 18-year boundary is known to be error-prone, so the claimed safety guarantee cannot be accepted without per-age-band accuracy and error analysis.
[§4.3 and Figure 4.1] The architecture as drawn and described does not implement the promised age-appropriate response modification. Section 4.3 claims that 'the secure query filter alters the responses to match the provided age category,' but in the public pipeline the age label is used only to decide whether the query passes through the Safe Query Filter; for users over 18 the query goes directly to the Fusion Hub, and for users under 18 it goes to the Safe Query Filter and then onward. Neither the Safe Query Filter (Section 4.4.2) nor any downstream module is described as rewriting or simplifying the final answer based on the detected age. The figure shows no path from the age-range detector to BVI-Friendly Prompt Management or to the response generator. Consequently, the safe-but-sensitive example in Section 4.2—a child receiving a simplified explanation of a mature exhibit—cannot be produced by the system as specified. The claim in the abstract and Section 1.2 that the system 'changes the responses to match audience less than 18 years' is therefore unsupported by the presented design.
[§3.3 and §5.4] The evaluation is not sufficient to support the central claim of an 'effective' or 'state-of-the-art' assistive system. The latency measurements come from a single hardware setup (Apple M2 in Section 3.3; Lenovo Y700 with GTX 960 in Section 5.4) with no reported variance, sample size, or statistical analysis for most figures. Figure 3.2 ('Evaluation of Cognition Engine across Different Query Lengths') and Figure 3.3 ('Impact of BVI-Friendly Response Generator') are presented without the underlying methodology: number of queries, composition of the test set, error bars, or significance tests. Figure 5.7 reports that MLLM latency 'appears to correlate' with output-word frequency from Google Ngram, but no correlation coefficient or p-value is given. Section 6.2 lists 'Evaluation of Multimodal Large Language Model Responses Based on Accepted Standards' and user studies as future work, which is an explicit admission that no formal accuracy, quality, or usability evaluation is contained in this manuscript. The effectiveness claim therefore remains a design hypothesis rather than a demonstrated result.
[§3.2.2.1 and §3.2.2.2] The latency-aware router is a distinctive contribution, but its behavior is not actually evaluated. The threshold value that triggers the switch from the multimodal Reasoning Engine to the Image-to-Text plus text-based path is described only as 'a certain threshold value'; no threshold is specified, and there is no sensitivity analysis showing how the end-to-end latency or answer quality changes as the threshold varies. Without such measurements, the claim that the router 'optimizes execution paths' and 'ensures uninterrupted communication' is not supported by evidence.

minor comments (5)

[§1.4] The text says 'Only 10% of BMI individuals can benefit from having a guide dog'; this appears to be a typo for 'BVI'.
[§5.2.4] The active interaction mode is triggered by keyboard spacebar input, which is not accessible to a blind user; the text acknowledges this as a prototype limitation, but it should be stated more prominently in the abstract or contributions to avoid overstating the prototype's accessibility.
[§2.1.3] The transformer description says the architecture 'was endorsed by Vaswani et al.'; the intended meaning is 'introduced by Vaswani et al.'
[Figure 3.3] The bar chart would be much more informative with the actual percentages and the number of test responses shown on or above the bars, and with a clear description of how 'inappropriate' was labeled.
[§5.4] The Ngram-based latency analysis is interesting but under-specified; please state how output vocabulary was mapped to Ngram frequencies, which Ngram corpus/version was used, and how many distinct output texts went into the correlation.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: Audo-Sight is a systems thesis whose components are implemented and measured; no prediction reduces to a fit or to a self-citation.

full rationale

The claimed contributions are architectural and empirical rather than derivational. The paper describes a private architecture (Ch. 3) whose components—Whisper STT, YOLO image-to-text, an MLLM (LLaVA-7B / Llama 3.2 11B Vision), NeMo Guardrails, GTTS—are externally sourced and evaluated with direct latency measurements in Ch. 5. The latency-aware LLM router uses a hand-set response-time threshold ('if the response time of the AI Reasoning Engine exceeds the set limit, the system redistributes the processing to the Image-to-Text module'); this is a design rule, not a parameter fitted to data and then renamed as a prediction. The BVI-Friendly Response Generator is implemented through NeMo Guardrails rules and its effect is directly measured as a reduction in inappropriate-answer categories, so the evaluation is not a self-definitional re-statement. The public-mode age filtering is described as a CNN-LSTM age-range detector and 'LLama Guard 3-1B' safe query filter, but the thesis gives no accuracy evaluation for the age classifier; that is an unvalidated safety claim, which is a correctness/validation risk, not a circularity. Future-work references to FastMig, UMS, and HEET (works by the advisor's group) are suggestions for possible optimization, not evidence used to justify a central result. No equation or fitted parameter is shown to reduce to another by construction, and no load-bearing argument rests on a self-citation. Therefore the circularity burden is not met: score 0.

Assumptions & free parameters 4 free parameters · 4 assumptions · 3 invented entities

The paper is a systems integration; the central claim does not rest on fitted parameters, but several components are introduced with no quantitative validation. The age detector and guardrail-based rewriting are the least supported pieces.

free parameters (4)

Latency routing threshold
A predefined response time limit in Section 3.2.2.1 that determines whether to use multimodal reasoning or the faster image-to-text and text reasoning path. The value is not reported, and the paper provides no sensitivity analysis.
Voice authentication distance threshold
A Euclidean distance threshold used in user identification (Section 3.2.1). Chosen by hand; no false acceptance or rejection rates are given.
Age cutoff = 18
The boundary between child-safe and adult content in the public mode (Section 4.3). A policy choice, not derived from evidence.
Object permanence parameter
Controls how long an object can leave the frame before it is restated in SmartSight's novelty prioritization (Section 5.3). Design choice, no tuning or evaluation.

assumptions (4)

domain assumption Voice-based age estimation from tone, pitch, and pronunciation can reliably separate under-18 from adult users.
Section 4.4.1 describes the CNN-LSTM age-range detector but provides no accuracy evaluation.
domain assumption NeMo Guardrails can rewrite LLM responses to be blind-friendly without introducing errors or losing information.
Section 3.2.2.7 describes the rewriting rules; no evaluation of factual preservation is provided.
domain assumption VizWiz-VQA is a representative benchmark for the target use case.
Section 3.3.2 selects VizWiz-VQA for evaluation, but the reported experiments only measure latency, not answer accuracy.
domain assumption The MLLM (Llama 3.2 11B Vision Instruct) provides accurate visual understanding.
The Reasoning Engine in Section 3.2.2.6 depends on this; no independent accuracy results are reported.

invented entities (3)

Audo-Sight system architecture
purpose: Provides multimodal, voice-based ambient interaction for BVI users in private and public settings
The architecture is the paper's contribution; there is no external validation or falsifiable handle beyond the described prototype.
BVI-Friendly Response Generator
purpose: Rewrites MLLM output to remove visual references and vague directions
Implemented with NeMo Guardrails; effectiveness is shown only in a self-reported bar chart (Figure 3.3).
Age-Range Detector
purpose: Estimates whether a user is under or over 18 from voice
No accuracy data or external benchmark is provided.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Audo-Sight: Enabling Ambient Interaction For Blind And Visually Impaired Individuals." pith.science (2026). https://pith.science/paper/GIQF7BZY

@misc{pith2026250500153,
  author       = {Pith},
  title        = {Pith review of: Audo-Sight: Enabling Ambient Interaction For Blind And Visually Impaired Individuals},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/GIQF7BZY}},
  note         = {Machine review of arXiv:2505.00153}
}

read the original abstract

Visually impaired people face significant challenges when attempting to interact with and understand complex environments, and traditional assistive technologies often struggle to quickly provide necessary contextual understanding and interactive intelligence. This thesis presents Audo-Sight, a state-of-the-art assistive system that seamlessly integrates Multimodal Large Language Models (MLLMs) to provide expedient, context-aware interactions for Blind and Visually Impaired (BVI) individuals. The system operates in two different modalities: personalized interaction through user identification and public access in common spaces like museums and shopping malls. In tailored environments, the system adjusts its output to conform to the preferences of individual users, thus enhancing accessibility through a user-aware form of interaction. In shared environments, Audo-Sight employs a shared architecture that adapts to its current user with no manual reconfiguration required. To facilitate appropriate interactions with the LLM, the public Audo-Sight solution includes an Age-Range Determiner and Safe Query Filter. Additionally, the system ensures that responses are respectful to BVI users through NeMo Guardrails. By utilizing multimodal reasoning, BVI-cognizant response editing, and safeguarding features, this work represents a major leap in AI-driven accessibility technology capable of increasing autonomy, safety, and interaction for people with visual impairments in social settings. Finally, we present the integration of Audo-Sight and SmartSight, which enables enhanced situational awareness for BVI individuals. This integration takes advantage of the real-time visual analysis of SmartSight, combined with the extensive reasoning and interactive capabilities of Audo-Sight, and goes beyond object identification to provide context-driven, voice-controlled assistance in dynamic environments.

Figures

Figures reproduced from arXiv: 2505.00153 by the authors.

**Figure 1.1.** Bird’s-eye view of SmartSight 1.0 The limitations of traditional assistive methods are that, they are passive in interaction, their reliance on text-based language models that make communication, and the need for voice-based, multimodal assistance to ensure natural, seamless communication. These limitations drive the development of Audo-Sight, a real-time, voice-driven, LLM-powered assistive framework for BVI indiv… view at source ↗

**Figure 1.2.** Assistive Tech in Action: Audo-Sight Enhancing Awareness For instance, a visually impaired individual traveling to a museum may need more than just standard technologies that provide audio versions of captions. They may want to explore specific locations, ask complex questions related to artifacts, or receive navigation assistance. Current methods do not offer this level of customized, multimodal engagement, and as … view at source ↗

**Figure 3.1.** Audo-Sight: Private Architecture Diagram the auditory and visual information are processed in time to provide an integrated assistive experience. The input management system takes care of two crucial types of inputs: voice input and image input. The procedure of voice input is made on the basis of oral questions or commands that come from the user, which are then understood by the system. Audo-Sight can be integrate… view at source ↗

Figures from the paper (10 more)

**Figure 3.2.** Figure 3.2: Evaluation of Cognition Engine across Different Query Lengths The graph shows that cognition engine latency is affected more by the complexity of the tasks and semantics rather than by the query length. This result aligns with the focus of 37 [PITH_FULL_IMAGE:figure…

**Figure 3.3.** Figure 3.3: Impact of BVI-Friendly Response Generator The bar graph shows the percentage of inappropriate answers before and after Nemo Guardrails implementation, evidencing its role in enhancing answer quality and safety. The greatest impact is the overwhelming reduction of err…

**Figure 4.1.** Figure 4.1: Audo-Sight: Public Architecture Diagram 4.4.1. Age-Range Detector The voice input from the user is sent to the immediate component called the AgeRange Detector. Essential part of Audo-Sight, the Age-Range Detector module estimates a user’s age depending on voice tra…

**Figure 5.1.** Figure 5.1: SmartSight Prototype 5.2.2. Image Capture and Preprocessing To acquire the images, the Raspberry Pi peripheral device launches the Raspberry Pi camera tool lib camera-still, which allows control of aperture, shutter speed, and image dimensions. While starting this pr…

**Figure 5.2.** Figure 5.2: Passive Ambient Perception Data Flow Diagram for SmartSight Passive mode allows the system to voluntarily describe aspects of the environment 49 [PITH_FULL_IMAGE:figures/full_fig_p058_5_2.png]

**Figure 5.3.** Figure 5.3: Active Ambient Interaction Execution Flow for SmartSight 5.3. Ambient Guidance Ambient Guidance is the fusion of active and passive perception. The fig 5.4 indicates that in order to have ambient perception of the environment we need to have both active perception an…

**Figure 5.4.** Figure 5.4: SmartSight guidance as a result of fusing passive ambient perception and active ambient interaction are output via text-to-speech. Thus, a list of objects from the previous identification are preserved for each frame for comparison with objects on the current frame.…

**Figure 5.5.** Figure 5.5: Latency visualization of MLLM, Guardrails, and speech to text The processing time of the MLLM can fluctuate, taking as long as 8 seconds or as 53 [PITH_FULL_IMAGE:figures/full_fig_p062_5_5.png]

**Figure 5.6.** Figure 5.6: MLLM Latency does not correlate to image complexity. As shown in figure 5.6, there is no visible correlation between the number of objects in an image, and the latency of the MLLM. This does not necessarily indicate that the image content has zero effect on the proce…

**Figure 5.7.** Figure 5.7: MLLM Latency appears to correlate with the popularity or external usage frequency of its output vocabulary. 55 [PITH_FULL_IMAGE:figures/full_fig_p064_5_7.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Evaluating the Impact of AI-Powered Audiovisual Personalization on Learner Emotion, Focus, and Learning Outcomes
cs.AI 2025-05 unverdicted novelty 4.0 of 10

The paper is a design and evaluation plan for an AI-generated personalized audiovisual study environment, with no empirical results reported.

Reference graph

Works this paper leans on

54 extracted references · 51 canonical work pages · cited by 1 Pith paper

[1]

Aishwarya Agrawal, Jiasen Lu, Stanislaw Antol, Margaret Mitchell, C Lawrence Zit- nick, Dhruv Batra, and Devi Parikh, Vqa: Visual question answering , arXiv preprint arXiv:1505.00468 (2016), 1–25

arXiv 2016
[2]

Ali Mokhtari, Md Abir Hossen, Pooyan Jamshidi, and Mohsen Amini Salehi, Felare: fair scheduling of machine learning applications on heterogeneous edge systems , arXiv (2022), 1–10

work page 2022
[3]

Balakrishnan, K

A. Balakrishnan, K. Ramana, G. Ashok, W. Viriyasitavat, S. Ahmad, and T. R. Gadekallu, Sonar glass—artificial vision: comprehensive design aspects of a synchro- nization protocol for vision based sensors , Measurement: Journal of the International Measurement Confederation (2023), 1–30

work page 2023
[4]

Florian Bordes, Richard Yuanzhe Pang, Anurag Ajay, Alexander C. Li, Adrien Bardes, Suzanne Petryk, Oscar Ma˜ nas, Zhiqiu Lin, Anas Mahmoud, Bargav Jayaraman, Mark Ibrahim, Melissa Hall, Yunyang Xiong, Jonathan Lebensold, Candace Ross, Srihari Jayakumar, Chuan Guo, Diane Bouchacourt, Haider Al-Tahan, Karthik Padthe, Vasu Sharma, Hu Xu, Xiaoqing Ellen Tan, ...

arXiv 2024
[5]

Thanawat Chanikaphon and Mohsen Amini Salehi, UMS: Live migration of container- ized services across autonomous computing systems , Proceedings of the IEEE Global Communications Conference (GLOBECOM ’23), IEEE, 2023, pp. 1–6

work page 2023
[6]

Marco Di Donato, Andrea Esposito, and Giuseppe Di Battista, A survey on wear- able technology: History, state-of-the-art and current challenges , IEEE Access 6 (2018), 48258–48263, https://ieeexplore.ieee.org/document/8597266. 61

work page arXiv 2018
[7]

Columbia Encyclopedia, Guide dog., Columbia University, 6th Edition (2021), 1

work page 2021
[8]

Envision Team, What’s next for 2025: The future of assistive technol- ogy & ai assistants , December 2024, https://www.letsenvision.com/blog/ future-assistive-technology

work page 2025

Show all 54 references

[9]

Be My Eyes, Be my eyes website , 2025, https://www.bemyeyes.com/

2025
[10]

Gladia Team, What is openai whisper? , 2025, https://www.gladia.io/blog/ what-is-openai-whisper

2025
[11]

avatar to person

Z. Guo, Z. Wang, and X. Jin, “avatar to person” (atp) virtual human social ability enhanced system for disabled people , Wireless Communications and Mobile Computing 2021 (2021), 10

2021
[12]

Yu Hao, Fan Yang, Hao Huang, Shuaihang Yuan, Sundeep Rangan, John Ross Rizzo, Yao Wang, and Yi Fang,A multi-modal foundation model to assist people with blindness and low vision in environmental interaction , Journal of Imaging 10 (2024), 1–15

2024
[13]

Haotian Liu, Chunyuan Li, Qingyang Wu, Yong Jae Lee, Ollama: Llava , 2025, https: //ollama.com/library/llava:7b

2025
[14]

Irfan Ullah Izaz Khan, Shah Khusro, Technology-assisted white cane: evaluation and future directions, PeerJ (2018), e6232

2018
[15]

Kosiedowski, A

M. Kosiedowski, A. Radziuk, P. Szymaniak, W. Kapsa, T. Rajtar, and M. Stroinski, On applying ambient intelligence to assist people with profound intellectual and multiple disabilities, 2020, pp. 895–914

2020
[16]

Lhanhsin, Vizwiz dataset , 2025, https://www.kaggle.com/datasets/lhanhsin/ vizwiz?select=test

2025
[17]

Yuri Lin, Jean-Baptiste Michel, Erez Lieberman Aiden, Jon Orwant, William Brock- man, and Slav Petrov, Proceedings of the 50th annual meeting of the association for computational linguistics volume 2: Demo papers , ACL ’12 (2012), 1–423

2012
[18]

Haotian Liu, Chunyuan Li, Qingyang Wu, and Yong Jae Lee, Visual instruction tuning, arXiv (2023), 1–25

2023
[19]

S. Luo et al., Single-pixel sensor array for precise identification systems using machine 62 learning, RSC Advances (2025), 7844–7854, https://pubs.rsc.org/en/content/ articlehtml/2025/ra/d5ra00167f

2025
[20]

Sorawit Manatura, Thanawat Chanikaphon, Chantana Chantrapornchai, and Mohsen Amini Salehi, FastMig: Leveraging FastFreeze to establish robust service liq- uidity in Cloud 2.0 , Proceedings of the 17th IEEE International Conference on Cloud Computing (IEEE Cloud ’24), IEEE, 202...

2024
[21]

Zain Merchant, Abrar Anwar, Emily Wang, Souti Chattopadhyay, and Jesse Thomason, Generating contextually-relevant navigation instructions for blind and low vision people , arXiv (2024), 1–6

2024
[22]

Mohamed Dhiaeddine Messaoudi, Bob-Antoine J Menelas, and Hamid Mcheick, Review of navigation assistive tools and technologies for the visually impaired , Sensors (Basel) (2022), 1–29

2022
[23]

Gray, William Brockman, The Google Books Team, Joseph P

Jean-Baptiste Michel, Yuan Kui Shen, Aviva Presser Aiden, Adrian Veres, Matthew K. Gray, William Brockman, The Google Books Team, Joseph P. Pickett, Dale Hoiberg, Dan Clancy, Peter Norvig, Jon Orwant, Steven Pinker, Martin A. Nowak, and Erez Lieberman Aiden, Quantitative analy...

2010
[24]

Minseo Kim, Wei Shu, and Mohsen Amini Salehi, He2c: A holistic approach for allo- cating latency-sensitive ai tasks across edge-cloud , arXiv (2024), 1–6

2024
[25]

Ali Mokhtari, Saeid Ghafouri, Pooyan Jamshidi, and Mohsen Amini Salehi, HEET: A performance measure to quantify heterogeneity in distributed computing systems , Pro- ceedings of the 17th IEEE/ACM International Conference on Utility and Cloud Com- puting (UCC 2024), IEEE, 2024,...

2024
[26]

Le Monde, Chatbots are like parrots: They repeat without understand- ing, 2024, https://www.lemonde.fr/en/economy/article/2024/10/07/ chatbots-are-like-parrots-they-repeat-without-understanding_6728523_ 19.html

2024
[27]

Nick Muscavage, Ai adds bespoke features to ready-made tools , December 2024. 63

2024
[28]

NVIDIA, Nemo guardrails documentation , 2025, https://docs.nvidia.com/nemo/ guardrails/latest/index.html

2025
[29]

php?g=1308742&p=9997824

NYU Libraries, What large language models are , 2024, https://guides.nyu.edu/c. php?g=1308742&p=9997824

2024
[30]

Carnegie Mellon University School of Computer Science, Gabriel: Edge computing framework, 2025, https://www.cmu.edu/scs/edgecomputing/software/gabriel. html

2025
[31]

Louis, Understanding the leading causes of blindness , Feb

Washington University School of Medicine in St. Louis, Understanding the leading causes of blindness , Feb. 2025, https://ophthalmology.wustl.edu/ understanding-the-leading-causes-of-blindness

2025
[32]

Gonzalez, M

Isaac Ong, Amjad Almahairi, Vincent Wu, Wei-Lin Chiang, Tianhao Wu, Joseph E. Gonzalez, M. Waleed Kadous, and Ion Stoica, Routellm: Learning to route llms with preference data, arXiv (2025), 1–16

2025
[33]

OpenAI, Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, and Ilge Akkaya, Gpt-4 technical report, arXiv (2023), 1–100

2023
[34]

Giorgio Presti, Dragan Ahmetovic, Mattia Ducci, Cristian Bernareggi, Luca Ludovico, Adriano Barate, Federico Avanzini, and Sergio Mascetti, Watchout: Obstacle sonifica- tion for people with visual impairment or blindness , ASSETS ’19 (2019), 1–12

2019
[35]

r-universe.dev/googleLanguageR/doc/text-to-speech.html

rOpenSci, googlelanguager: Text-to-speech documentation , 2025, https://ropensci. r-universe.dev/googleLanguageR/doc/text-to-speech.html

2025
[36]

Kathrin Seßler, Arne Bewersdorff, Claudia Nerdel, and Enkelejda Kasneci, Towards adaptive feedback with ai: Comparing the feedback quality of llms and teachers on ex- perimentation protocols, arXiv preprint arXiv:2502.12842 (2023), 1–23

2023 arXiv
[37]

SHG Technologies, Aura vision glasses , 2025, https://shgtechnologies.com/ products/aura-vision-glasses

2025
[38]

Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, and Johan Schalkwyk et al., Gemini: A family of highly capable mul- timodal models, arXiv (2023), 1–90. 64

2023
[39]

OrCam Technologies, Orcam myeye: Ai-powered assistive device for the visually im- paired, 2023, https://www.orcam.com/en/myeye

2023
[40]

The Hive AI, Meta llama 3.2 11b vision instruct model , 2025, https://thehive.ai/ models/meta-llama/llama-3.2-11b-vision-instruct

2025
[41]

Gang Luo Unxian He, Shrinivas Pundlik, Can chatgpt assist visually impaired people with micro-navigation?, arXiv (2024), 1–15

2024
[42]

Ashish Vaswani, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Lukasz Kaiser, and Illia Polosukhin, Attention is all you need , Advances in Neural Information Processing Systems 30 (2017), 1–15

2017
[43]

VisionAid, About smart vision glasses , 2025, https://visionaid.org/ about-smart-vision-glasses/

2025
[44]

Hao Wang, Jiayou Qin, Ashish Bastola, Xiwen Chen, John Suchanek, Zihao Gong, and Abolfazl Razi, Visiongpt: Llm-assisted real-time anomaly detection for safe visual navigation, arXiv (2024), 1–15

2024
[45]

Wikipedia Contributors, Ai trust paradox , 2024, https://en.wikipedia.org/wiki/ AI_trust_paradox

2024
[46]

Iris Xie, Wonchan Choi, Shengang Wang, Hyun Seung Lee, Bo Hyun Hong, Ning-Chiao Wang, and Emmanuel Kwame Cudjoe, Help-seeking situations related to visual inter- actions on mobile platforms and recommended designs for blind and visually impaired users, Journal of Imaging 10 (2...

2024
[47]

Carroll, Emerging practices for large multimodal model (lmm) assistance for people with visual impairments: Implications for design , arXiv (2024), 1–22

Jingyi Xie, Rui Yu, He Zhang, Sooyeon Lee, Syed Masum Billah, and John M. Carroll, Emerging practices for large multimodal model (lmm) assistance for people with visual impairments: Implications for design , arXiv (2024), 1–22

2024
[48]

Sun, Jiannan Lu, Hadas Kotek, Yuhan Zhang, Paul McCarthy, Christopher Klein, Stephen Pulman, and Jason D

Zidi Xiu, Kai-Chen Cheng, David Q. Sun, Jiannan Lu, Hadas Kotek, Yuhan Zhang, Paul McCarthy, Christopher Klein, Stephen Pulman, and Jason D. Williams, Feedback effect in user interaction with intelligent assistants: Delayed engagement, adaption, and drop-out, arXiv preprint ar...

2023 arXiv
[49]

Bufang Yang, Lixing He, Kaiwei Liu, and Zhenyu Yan, Viassist: Adapting multi-modal large language models for users with visual impairments , arXiv (2024), 1–6

2024
[50]

, National Science Review 11 (2024), no

Shukang Yin, Chaoyou Fu, Sirui Zhao, Ke Li, Xing Sun, Tong Xu, and Enhong Chen, A survey on multimodal large language models. , National Science Review 11 (2024), no. 12, 1 – 20

2024
[51]

Yue YZheng, Yuhao Chen, Bin Qian, Xiufang Shi, Yuanchao Shu, and Jiming Chen, A review on edge large language models: Design, execution, and applications , arXiv (2024), 1–36

2024
[52]

Zuhair Dawood Zaghlool, Incorporating the impacts and limitations of ai-driven feedback mechanisms in foreign language teaching , 2023, https://migrationletters

Dr. Zuhair Dawood Zaghlool, Incorporating the impacts and limitations of ai-driven feedback mechanisms in foreign language teaching , 2023, https://migrationletters. com/index.php/ml/article/view/4863

2023
[53]

Yi Zhao, Yilin Zhang, Rong Xiang, Jing Li, and Hillming Li, Vialm: A survey and benchmark of visually impaired assistance with large models , arXiv (2024), 1–9

2024
[54]

SM Zobaed, Ali Mokhtari, Jaya Prakash Champati, Mathieu Kourouma, and Mohsen Amini Salehi, Edge-MultiAI: Multi-Tenancy of Latency-Sensitive Deep Learning Applications on Edge , arXiv (2022), 1–10. 66

2022

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.

[1] [1]

Aishwarya Agrawal, Jiasen Lu, Stanislaw Antol, Margaret Mitchell, C Lawrence Zit- nick, Dhruv Batra, and Devi Parikh, Vqa: Visual question answering , arXiv preprint arXiv:1505.00468 (2016), 1–25

arXiv 2016

[2] [2]

Ali Mokhtari, Md Abir Hossen, Pooyan Jamshidi, and Mohsen Amini Salehi, Felare: fair scheduling of machine learning applications on heterogeneous edge systems , arXiv (2022), 1–10

work page 2022

[3] [3]

Balakrishnan, K

A. Balakrishnan, K. Ramana, G. Ashok, W. Viriyasitavat, S. Ahmad, and T. R. Gadekallu, Sonar glass—artificial vision: comprehensive design aspects of a synchro- nization protocol for vision based sensors , Measurement: Journal of the International Measurement Confederation (2023), 1–30

work page 2023

[4] [4]

Florian Bordes, Richard Yuanzhe Pang, Anurag Ajay, Alexander C. Li, Adrien Bardes, Suzanne Petryk, Oscar Ma˜ nas, Zhiqiu Lin, Anas Mahmoud, Bargav Jayaraman, Mark Ibrahim, Melissa Hall, Yunyang Xiong, Jonathan Lebensold, Candace Ross, Srihari Jayakumar, Chuan Guo, Diane Bouchacourt, Haider Al-Tahan, Karthik Padthe, Vasu Sharma, Hu Xu, Xiaoqing Ellen Tan, ...

arXiv 2024

[5] [5]

Thanawat Chanikaphon and Mohsen Amini Salehi, UMS: Live migration of container- ized services across autonomous computing systems , Proceedings of the IEEE Global Communications Conference (GLOBECOM ’23), IEEE, 2023, pp. 1–6

work page 2023

[6] [6]

Marco Di Donato, Andrea Esposito, and Giuseppe Di Battista, A survey on wear- able technology: History, state-of-the-art and current challenges , IEEE Access 6 (2018), 48258–48263, https://ieeexplore.ieee.org/document/8597266. 61

work page arXiv 2018

[7] [7]

Columbia Encyclopedia, Guide dog., Columbia University, 6th Edition (2021), 1

work page 2021

[8] [8]

Envision Team, What’s next for 2025: The future of assistive technol- ogy & ai assistants , December 2024, https://www.letsenvision.com/blog/ future-assistive-technology

work page 2025