Pith. sign in

REVIEW 5 major objections 4 minor 37 references

HyperCam: Low-Power Onboard Computer Vision for IoT Cameras

T0 review · 5 major / 4 minor · reviewed 2026-08-10 · deepseek-v4-flash

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

Pith's one-line read HyperCam shows that hyperdimensional computing can run image classification entirely on a low-power IoT camera, using under 63 KB of flash and 22.25 KB of RAM while matching or beating tiny neural networks on several vision tasks.

desk verdict Credible systems paper with a genuinely new sparse-bundling encoding; the resource numbers hold up, the accuracy claims need error bars and a held-out d. read the letter →

arxiv 2501.10547 v1 pith:HSB2STIH submitted 2025-01-17 cs.CV cs.LGcs.NEeess.IV

Chae Young Lee , Pu (Luke) Yi , Maxwell Fite , Tejus Rao , Sara Achour , Zerina Kapetanovic This is my paper · ORCID
classification cs.CVcs.LGcs.NEeess.IV
keywords hyperdimensionalcomputingvectorsymbolicarchitecturesIoTcamerasembeddedmachinelearningimageclassificationlow-powerinferencesparsebundlingmicrocontroller
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

This paper introduces HyperCam, an image-classification pipeline that runs entirely on a microcontroller inside a low-power wireless camera. It argues that hyperdimensional computing (HDC), which represents images as high-dimensional binary vectors, can replace deep neural networks for onboard vision, and that a new sparse bundling operator makes HDC encoding fast enough and small enough for commodity MCUs. On MNIST, Fashion-MNIST, face detection, and face identification, HyperCam reports accuracies of 93.60%, 84.06%, 92.98%, and 72.79% while using 42.91–63.00 KB of flash, 22.25 KB of RAM, and 0.08–0.27 s inference latency. Compared with SVM, xgBoost, MicroNets, MobileNetV3, and MCUNetV3, the paper claims HyperCam is the only classifier that is both accurate enough and resource-efficient enough for low-power camera systems. A sympathetic reader would care because, if correct, this makes privacy-preserving, offline computer vision practical on tiny battery-powered devices.

What carries the argument

The load-bearing object is the sparse bundling operator (Algorithm 1), which approximates the standard HDC bundling of a set of permuted hypervectors. Given a set of integer pixel positions, it hashes each element into d=20 positions of a 10,000-bit hypervector, either setting bits (Bloom Filter backend) or adding ±1 signed increments and then thresholding at zero (Count Sketch backend). This replaces O(n)=10,000 bit operations per bundle with O(d)=20, a 500-fold reduction, while preserving the key property that similar sets produce nearby hypervectors. Around it, the encoder uses permutation-based codebooks so no position hypervectors are stored, level-based value hypervectors generated on the fly, and weighted bundling to restore pixel-value frequency lost in factoring. Together these make the per-image cost 19,200 sparse bundling operations plus 256 dense bundling operations instead of 19,200 full bundling operations.

What would settle it

Retrain HyperCam on a task with more classes or higher-resolution images (for example, 10-class or 32x32 CIFAR-10, or the 8-class face dataset upscaled) and compare accuracy, memory, and latency against the exact Rewrite2 encoder; if the accuracy gap grows well beyond the 1-point MNIST drop reported here while exact bundling holds, then the sparse approximation, not the HDC representation, is the accuracy bottleneck.

Watch

Extended reading notes

Core claim

The central claim is that a carefully rewritten HDC encoder can classify images on a microcontroller with accuracy close to neural networks but with a fraction of the memory and latency. The paper shows this by starting from a naive pixel-based HDC encoding and applying four rewrites: permutation-based position codebooks, coalescing row and column indices into a single codebook, factoring value hypervectors out of the bundling with a weighted re-introduction of pixel-value frequency, and finally replacing dense bundling with a sparse bundling operator built on a Bloom Filter or Count Sketch. The resulting encoding uses two orders of magnitude fewer bundling operations, cutting the codebook from 536 stored hypervectors to two on-the-fly generated ones. On the hardware prototype, HyperCam's Count Sketch version achieves 93.60% MNIST accuracy at 63.00 KB flash and 0.26 s latency, and the Bloom Filter version cuts latency to 0.08 s at 52.62 KB flash with a 3.24-point MNIST accuracy drop. The paper's claim is that this trade-off is the right operating point for IoT cameras, where energy, not peak accuracy, is the binding constraint.

Load-bearing premise

The sparse bundling operator preserves enough of the distance structure of full hyperdimensional bundling that learned class hypervectors remain separable, even though no error bound or approximation analysis is provided.

Editorial extensions

If this is right

  • HyperCam can be ported to any MCU with bitwise operations, needing no floating-point units, NPUs, or ML libraries.
  • Adding a new class to the classifier costs only n=10,000 bits of flash, so models can be extended in the field.
  • Onboard inference removes the need to transmit raw images, reducing communication overhead and privacy exposure for IoT cameras.
  • The reported latency of 0.08–0.27 s supports real-time event-triggered sensing on battery power.
  • The Bloom Filter backend offers an explicit accuracy-versus-latency knob for deployments.

Reading between the lines

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

  • The same sparse bundling operator could accelerate other HDC workloads beyond vision, such as sensor fusion or sequence encoding, wherever bundling over sets of permuted hypervectors dominates cost.
  • The density parameter d=20 is selected experimentally; a theoretical analysis of how d scales with hypervector dimension and set size could turn HyperCam into a tunable system rather than a fixed configuration.
  • Because the encoder loses accuracy on high-resolution or many-class inputs, a hybrid design could use sparse bundling for an initial pass and exact bundling only for hard examples.
  • The paper's comparison omits the energy cost of the camera sensor itself; a full system-level energy model including capture and transmission would let one test whether the 0.08–0.27 s inference window is the dominant power term.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

5 major / 4 minor

Summary. The paper presents HyperCam, an HDC-based image classification pipeline for low-power IoT cameras. It introduces four encoding rewrites to reduce codebook memory, binding, and bundling operations, culminating in a sparse bundling operator based on Bloom filters and Count Sketches. The authors implement the pipeline on an STM32U585AI MCU with a Himax camera and BLE, and report accuracy, flash, RAM, latency, and power across MNIST, Fashion-MNIST, and custom face detection/identification tasks. They claim HyperCam is the only classifier that combines competitive accuracy with resource usage suitable for low-power camera systems.

Significance. If the measurements are robust, HyperCam provides a compelling data point that HDC image classification can run on commodity MCUs with 53-63 KB flash, 22.25 KB RAM, and under 0.3 s latency while retaining useful accuracy on simple vision tasks. The detailed algorithm pseudocode, per-stage latency profiling, flash-memory breakdown, and real hardware power measurements are valuable for the embedded ML and IoT camera communities. However, the significance is currently limited by the lack of statistical error bars, the unanalyzed and task-tuned sparse bundling approximation, and a few inconsistencies in the reported comparison.

major comments (5)
  1. [Section 6.1.3, Table 2] Accuracy is reported as a single point estimate from one 80/20 split, without standard deviations, confidence intervals, or multiple random seeds. This matters because HyperCam's encoding uses randomly generated codebooks and OnlineHD training is stochastic, and several comparisons in Table 2 are very close (e.g., HyperCam* 92.98% vs. MicroNets 92.86% on Face Detection; HyperCam* 84.06% vs. MobileNetV3 86.48% on Fashion-MNIST). With no variance estimate, the abstract's claim that HyperCam is 'the only classifier that achieves competitive accuracy while maintaining competitive memory footprint and inference latency' is not statistically supported. Please report mean and standard deviation over multiple splits and seeds and, where possible, a paired significance test.
  2. [Section 4.2.4 and Section 4.3] The sparse bundling approximation is the only semantics-breaking rewrite that affects all reported accuracies, yet no approximation guarantee or distance-preservation analysis is provided. The density parameter d=20 is selected experimentally on the same four evaluation tasks, and the paper states that d<20 leads to a sharp accuracy drop, but no held-out task, resolution, class-count sweep, or error bound is reported. This is load-bearing: Table 2 shows that Rewrite 4 costs 5.84 accuracy points on Face Identification (Rewrite 2 78.63% to HyperCam* 72.79%), the largest drop in the table and exactly the task used in the abstract headline. Please provide either a theoretical bound on the distance distortion introduced by SparseBundle, or a systematic sensitivity analysis on tasks and distributions not used to choose d, before the claimed generality can be evaluated.
  3. [Algorithm 1, lines 15-21] The Count Sketch finalization in Algorithm 1 as written is `if not bloom then for i in 0..n: hv[i] = 1 ? hv[i] >= 0 : 0`, i.e., every non-negative entry (including all zero entries) is mapped to 1. For a singleton set this yields a vector with about n - d/2 ones rather than d ones, contradicting the claimed sparse density and the O(d) complexity argument; for large sets the output density saturates differently than exact HDC bundling. If the actual implementation thresholds at `hv[i] > 0` or uses a different rule, the pseudocode must be corrected and the density behavior analyzed. The dependence of output density on set size is not accounted for in Eq. (1), where sparse bundles for different pixel values are weighted and combined before a global threshold; the paper should explain why this structural mismatch does not bias classes with different pixel-value histograms.
  4. [Abstract and Section 6.2] The headline claim that HyperCam is 'the only classifier that achieves competitive accuracy while maintaining competitive memory footprint and inference latency' is not operationalized. In Table 2, MCUNetV3* achieves higher accuracy on all four tasks (e.g., 99.34% vs. 93.60% on MNIST) at 1190 KB flash and 6.7 s latency, and xgBoost achieves 94.46% vs. 92.98% on Face Detection at 134.92 KB flash and 0.01 s latency. Without a stated definition of 'competitive accuracy' (e.g., within X percentage points of the best model) and a stated resource envelope, the 'only' claim cannot be evaluated or falsified. Please state the accuracy threshold and resource budget used to justify the claim, or weaken the claim accordingly.
  5. [Table 2, OnlineHD row] The OnlineHD row reports the same accuracy, 84.62%, for both Face Detection and Face Identification. If this is a typo, it should be corrected; if it is not, the paper should explain why the 7-class identification accuracy exactly equals the binary detection accuracy. As printed, this number also undermines the text's assertion that HD classifiers including HyperCam show a 'graceful decline' to 72.79% on Face Identification, because a baseline HDC method is listed at 84.62% on the same task. Please verify this entry and clarify the comparison.
minor comments (4)
  1. [Section 5.3] The microcontroller is referred to as 'STM32UF855AI' in Section 5.3 but as 'STM32U585AI' elsewhere; please correct the typo.
  2. [Section 6.2] The text says 'HyperCam's most memory-efficient version (count-sketch) requires only 63 KB of flash memory,' but Table 2 shows that the Bloom Filter version (HyperCam**) uses 52.62 KB, which is smaller than the Count Sketch version's 63.00 KB; this appears to be a swapped label and should be fixed.
  3. [Section 6.3 and Table 4] The active power is reported as 128 mW during image processing and as 102 mW average over the 450 ms active period in Section 6.3, while Table 4 lists Active Power as 128 mW; please clarify which quantity is being reported and make the table consistent with the text.
  4. [Contributions and Section 8] The paper states that HyperCam code is open source, but no repository URL or access instructions are provided; please add a link to the code to support reproducibility.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: HyperCam's accuracy claims are empirical benchmark results, not derivations that assume their own conclusions.

full rationale

HyperCam's central claims are evaluated externally rather than derived from fitted inputs. The accuracy numbers in Table 2 are measured on held-out test splits (8:2 split, Section 6.1.3), and the sparse-bundling density d=20 is an experimental configuration reported as selected by accuracy behavior, not a parameter fitted to reproduce the table's results. The paper explicitly identifies Rewrite 4 as semantics-breaking and reports the accuracy loss versus Rewrite 2, so the approximation is not disguised as an exact derivation. Self-citations (WISPCam, FarmBeats) appear only in related-work context and are not load-bearing; no uniqueness theorem or prior result by the same authors is invoked to force the design. The lack of an error bound for the sparse-bundling approximation is a generalization and soundness concern, not circularity: the approximation is evaluated by its measured accuracy on external datasets rather than defined to match those accuracies. The Rewrite 2 baseline is an internal comparison point, but the final HyperCam accuracy is independently benchmarked against external datasets and external baselines. No equation or claim in the paper reduces by construction to its own inputs.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central result rests on standard HDC assumptions about random hypervector orthogonality and operator behavior, on the ad hoc assumption that sparse hash-based bundling preserves enough structure at d = 20, and on a statistical assumption that a single split gives a stable accuracy estimate. The only tuned numeric parameters are n, d, and OnlineHD hyperparameters, with d being the most consequential because it controls the lossy approximation. No new physical entities are introduced.

free parameters (3)
  • Sparse bundling density d = 20
    Set experimentally in Section 4.2.4; the paper reports a sharp accuracy drop for d < 20. It is chosen on the same tasks used for final accuracy, so it may absorb task-specific structure.
  • Hypervector dimension n = 10,000
    Chosen hypervector length used for all experiments in Section 6.1.2. It is a standard HDC choice, but no ablation or sensitivity analysis is reported.
  • OnlineHD training hyperparameters = not reported
    Section 3 says training uses a binarized modification of OnlineHD, but the number of passes, update rates, and convergence criteria are not given. These affect the item memory and final accuracy.
assumptions (5)
  • domain assumption Randomly generated binary hypervectors are near orthogonal, and XOR binding, majority bundling, and permutation preserve the similarity structure needed for classification.
    Standard BSC-HDC assumption invoked throughout Section 2 and used implicitly in Rewrites 1 through 4 in Section 4.2.
  • ad hoc to paper Sparse bundling with Bloom Filter and Count Sketch approximates full HDC bundling well enough at d = 20 for image classification.
    Central to Section 4.2.4 and Algorithm 1. No error bound or approximation analysis is given, and d is chosen experimentally.
  • domain assumption The binarized modification of OnlineHD retains the convergence and accuracy properties of the real-valued OnlineHD algorithm.
    Section 3 states HyperCam uses a modified OnlineHD that binarizes classifier vectors after each iteration, but no proof or analysis of the modification is provided.
  • domain assumption Post-training integer quantization of SVM, XGBoost, MicroNets, MobileNetV3, and MCUNetV3 does not unfairly disadvantage them relative to HyperCam.
    Section 6.1.2 states all non-HDC models are quantized post-training. Quantization sensitivity can vary by model, and no quantization-aware training is reported.
  • domain assumption One 80/20 train/test split per dataset yields a reliable accuracy estimate.
    Section 6.1.3 defines accuracy from a single split. No cross-validation, repeated runs, or confidence intervals are reported.

how reviews work

0 comments
Cite this review

Pith. "Pith review of HyperCam: Low-Power Onboard Computer Vision for IoT Cameras." pith.science (2026). https://pith.science/paper/HSB2STIH

@misc{pith2026250110547,
  author       = {Pith},
  title        = {Pith review of: HyperCam: Low-Power Onboard Computer Vision for IoT Cameras},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/HSB2STIH}},
  note         = {Machine review of arXiv:2501.10547}
}
read the original abstract

We present HyperCam, an energy-efficient image classification pipeline that enables computer vision tasks onboard low-power IoT camera systems. HyperCam leverages hyperdimensional computing to perform training and inference efficiently on low-power microcontrollers. We implement a low-power wireless camera platform using off-the-shelf hardware and demonstrate that HyperCam can achieve an accuracy of 93.60%, 84.06%, 92.98%, and 72.79% for MNIST, Fashion-MNIST, Face Detection, and Face Identification tasks, respectively, while significantly outperforming other classifiers in resource efficiency. Specifically, it delivers inference latency of 0.08-0.27s while using 42.91-63.00KB flash memory and 22.25KB RAM at peak. Among other machine learning classifiers such as SVM, xgBoost, MicroNets, MobileNetV3, and MCUNetV3, HyperCam is the only classifier that achieves competitive accuracy while maintaining competitive memory footprint and inference latency that meets the resource requirements of low-power camera systems.

Figures

Figures reproduced from arXiv: 2501.10547 by the authors.

Figure 1
Figure 1. HDC for image classification. HyperCam uses an HD classifier to perform face detection and identification tasks onboard low-power wireless camera platforms. example, in data-driven agriculture, embedded ML can an￾alyze image data to assess crop yields or detect pests and plant diseases [33]. It can also support environmental and wildlife monitoring through camera traps or field survey robots [1, 2]. In these scenari… view at source ↗
Figure 2
Figure 2. Memory layout of STM32U585AI. and meeting these constraints requires careful model de￾sign. Additionally, optimizing latency is critical not only to meet real-time requirements but also to minimize the overall power consumption of the system. In image processing, the HD computation load increases proportionally to the image size. For example, a baseline HDC approach can take one minute to classify a 120 × 160 graysc… view at source ↗
Figure 3
Figure 3. Key operations of BSC. (1) Basis vectors are generated for every letter. (2) Binding of data creates a record. (3) Bundling of words creates a set. (4) Permutation is applied to create hypervectors on-the-fly. 2.1.4 Permutation. The permutation operator (𝑝) is imple￾mented as a circular shift, which creates a dissimilar hyper￾vector far apart from the input. Because of this characteristic, permutation is used to cre… view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: HyperCam overview. The HyperCam classifier runs onboard a low-power wireless camera platform and has three key components: image encoder, training algorithm, and inference algorithm. binarized item memory to find misclassifications and update the real-valued model. Inf…
Figure 5
Figure 5. Figure 5: Sample images in the collected dataset. 5.2 Data Collection In IoT deployments, collecting data from real-world scenarios is crucial for aligning with actual conditions. IoT systems are sensitive to environmental changes, sensor noise, and operat￾ing conditions (e.g., …
Figure 8
Figure 8. Figure 8: Latency Profiling of HyperCam. the highest accuracy levels (e.g., 98.69% on MNIST for Mo￾bileNetV3) but at the cost of substantially higher memory usage and latency. For example, in terms of flash memory consumption, MicroNets requires over 500 KB of flash mem￾ory whil…
Figure 7
Figure 7. Figure 7: Breakdown of flash memory. 0 40 80 120 160 200 240 Time (ms) Count Sketch Bloom Filter Bind & Sum Binning Initialization Majority Vote Prediction Encoder Count Sketch Bloom Filter [PITH_FULL_IMAGE:figures/full_fig_p010_7.png]
Figure 9
Figure 9. Figure 9: System Power Consumption. (1) camera plat￾form in sleep mode (2) camera initialization, image capture, and inference, (3) data transmission, and (4) system returns to sleep mode. On the other hand, there have been works exploring the use of hyperdimensional computing f…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

37 extracted references · 27 canonical work pages

  1. [1]

    Sayed Saad Afzal, Waleed Akbar, Osvy Rodriguez, Mario Doumet, Unsoo Ha, Reza Ghaffarivardavagh, and Fadel Adib. 2022. Battery-free wireless imaging of underwater environments. Nature Communica- tions 13 (2022), 5546. https://doi.org/10.1038/s41467-022-33223-x

    work page doi:10.1038/s41467-022-33223-x 2022

  2. [2]

    Jorge A Ahumada, Eric Fegraus, Tanya Birch, Nicole Flores, Roland Kays, Timothy G O’Brien, Jonathan Palmer, Stephanie Schuttler, Jen- nifer Y Zhao, Walter Jetz, and et al. 2020. Wildlife Insights: A Platform to Maximize the Potential of Camera Trap and Other Passive Sensor Wildlife Data for the Planet. Environmental Conservation 47, 1 (2020), 1–6. https:/...

    doi:10.1017/s0376892919000298 2020

  3. [3]

    Banbury, Chuteng Zhou, Igor Fedorov, Ramon Matas Navarro, Urmish Thakker, Dibakar Gope, Vijay Janapa Reddi, Matthew Mat- tina, and Paul N

    Colby R. Banbury, Chuteng Zhou, Igor Fedorov, Ramon Matas Navarro, Urmish Thakker, Dibakar Gope, Vijay Janapa Reddi, Matthew Mat- tina, and Paul N. Whatmough. 2020. MicroNets: Neural Network Architectures for Deploying TinyML Applications on Commodity Microcontrollers. CoRR abs/2010.11267 (2020). arXiv:2010.11267 12 HyperCam: Low-Power Onboard Computer Vi...

    arXiv 2020

  4. [4]

    Ron Banner, Itay Hubara, Elad Hoffer, and Daniel Soudry. 2018. Scal- able methods for 8-bit training of neural networks. In Proceedings of the 32nd International Conference on Neural Information Processing Systems (Montréal, Canada) (NIPS’18). Curran Associates Inc., Red Hook, NY, USA, 5151–5159

    work page 2018

  5. [5]

    Burton H Bloom. 1970. Space/time trade-offs in hash coding with allowable errors. Commun. ACM 13, 7 (1970), 422–426

    1970

  6. [6]

    Moses Charikar, Kevin Chen, and Martin Farach-Colton. 2002. Find- ing frequent items in data streams. In International Colloquium on Automata, Languages, and Programming. Springer, 693–703

    work page 2002

  7. [7]

    Kenneth L Clarkson, Shashanka Ubaru, and Elizabeth Yang. 2023. Capacity analysis of vector symbolic architectures. arXiv preprint arXiv:2301.10352 (2023)

    work page Pith review arXiv 2023

  8. [8]

    Robert David, Jared Duke, Advait Jain, Vijay Janapa Reddi, Nat Jeffries, Jian Li, Nick Kreeger, Ian Nappier, Meghna Natraj, Tiezhen Wang, et al. 2021. Tensorflow lite micro: Embedded machine learning for tinyml systems. Proceedings of Machine Learning and Systems 3 (2021), 800–811

    work page 2021

Show all 37 references
  1. [9]

    Alejandro Hernández-Cano, Namiko Matsumoto, Eric Ping, and Mohsen Imani. 2021. OnlineHD: Robust, Efficient, and Single-Pass Online Learning Using Hyperdimensional System. In 2021 Design, Automation & Test in Europe Conference & Exhibition (DATE) . 56–61. https://doi.org/10.239...

    2021

  2. [10]

    Himax. 2024. HM01B0 Ultralow Power CIS. https: //www.himax.com.tw/products/cmos-image-sensor/always-on- vision-sensors/hm01b0/

    2024

  3. [11]

    Mohsen Imani, Abbas Rahimi, Deqian Kong, Tajana Rosing, and Jan M Rabaey. 2017. Exploring hyperdimensional associative memory. In 2017 IEEE International Symposium on High Performance Computer Architecture (HPCA). IEEE, 445–456. https://doi.org/10.1109/HPCA. 2017.28

    2017 doi

  4. [12]

    Mohsen Imani, Ali Zakeri, Hanning Chen, TaeHyun Kim, Prathyush Poduval, Hyunsei Lee, Yeseong Kim, Elaheh Sadredini, and Farhad Imani. 2022. Neural computation for robust and holographic face detection. In Proceedings of the 59th ACM/IEEE Design Automation Conference (San Franc...

    2022

  5. [13]

    Vikram Iyer, Ali Najafi, Johannes James, Sawyer Fuller, and Shyamnath Gollakota. 2020. Wireless steerable vision for live insects and insect- scale robots. Science robotics 5, 44 (2020), eabb0839

    2020

  6. [14]

    Benoit Jacob, Skirmantas Kligys, Bo Chen, Menglong Zhu, Matthew Tang, Andrew Howard, Hartwig Adam, and Dmitry Kalenichenko

  7. [15]

    Colleen Josephson, Lei Yang, Pengyu Zhang, and Sachin Katti. 2019. Wireless computer vision using commodity radios. In Proceedings of the 18th International Conference on Information Processing in Sensor Networks. 229–240

    2019

  8. [16]

    Joulescope. 2024. Joulescope JS220: Precision Energy Ana- lyzer. https://www.joulescope.com/products/js220-joulescope- precision-energy-analyzer

    2024

  9. [17]

    Pentti Kanerva. 1997. Fully distributed representation. PAT 1, 5 (1997), 10000

    1997

  10. [18]

    Denis Kleyko, Mike Davies, Edward Paxon Frady, Pentti Kanerva, Spencer J Kent, Bruno A Olshausen, Evgeny Osipov, Jan M Rabaey, Dmitri A Rachkovskij, Abbas Rahimi, et al . 2022. Vector symbolic architectures as a computing framework for emerging hardware. Proc. IEEE 110, 10 (20...

    2022

  11. [19]

    Kleyko, D

    D. Kleyko, D. A. Rachkovskij, E. Osipov, and A. Rahimi. 2022. A Survey on Hyperdimensional Computing Aka Vector Symbolic Architectures, Part I: Models and Data Transformations. ACM Comput. Surv. 55, 6, Article 130 (dec 2022), 40 pages. https://doi.org/10.1145/3538531

    2022 doi

  12. [20]

    Denis Kleyko, Abbas Rahimi, Ross W Gayler, and Evgeny Osipov. 2020. Autoscaling bloom filter: controlling trade-off between true and false positives. Neural Computing and Applications 32 (2020), 3675–3684

    2020

  13. [21]

    Jovin Langenegger, Geethan Karunaratne, Michael Hersche, Luca Benini, Abu Sebastian, and Abbas Rahimi. 2023. In-memory factoriza- tion of holographic perceptual representations.Nature Nanotechnology 18, 5 (2023), 479–485

    2023

  14. [22]

    Yann LeCun, Corinna Cortes, and Christopher J.C. Burges. 1998. The MNIST Database of Handwritten Digits. http://yann.lecun.com/exdb/ mnist/

    1998

  15. [23]

    Ji Lin, Wei-Ming Chen, Yujun Lin, John Cohn, Chuang Gan, and Song Han. 2020. MCUNet: tiny deep learning on IoT devices. In Proceedings of the 34th International Conference on Neural Information Processing Systems (Vancouver, BC, Canada) (NIPS’20). Curran Associates Inc., Red H...

    2020

  16. [24]

    Ji Lin, Wei-Ming Chen, Yujun Lin, Chuang Gan, Song Han, et al. 2020. Mcunet: Tiny deep learning on iot devices. Advances in Neural Infor- mation Processing Systems 33 (2020), 11711–11722

    2020

  17. [25]

    Ji Lin, Ligeng Zhu, Wei-Ming Chen, Wei-Chen Wang, Chuang Gan, and Song Han. 2022. On-device training under 256kb memory. Advances in Neural Information Processing Systems 35 (2022), 22941–22954

    2022

  18. [26]

    Erez Manor and Shlomo Greenberg. 2022. Custom hardware inference accelerator for tensorflow lite for microcontrollers. IEEE Access 10 (2022), 73484–73493

    2022

  19. [27]

    Saman Naderiparizi, Zerina Kapetanovic, and Joshua R Smith. 2016. Wispcam: An rf-powered smart camera for machine vision appli- cations. In Proceedings of the 4th International Workshop on Energy Harvesting and Energy-Neutral Sensing Systems . 19–22

    2016

  20. [28]

    Saman Naderiparizi, Aaron N Parks, Zerina Kapetanovic, Benjamin Ransford, and Joshua R Smith. 2015. WISPCam: A battery-free RFID camera. In 2015 IEEE International Conference on RFID (RFID) . IEEE, 166–173

    2015

  21. [29]

    Dmitriy A Rachkovskiy, Sergey V Slipchenko, Ernst M Kussul, and Tatyana N Baidyk. 2005. Sparse binary distributed encoding of scalars. Journal of Automation and Information Sciences 37, 6 (2005)

    2005

  22. [30]

    Nordic Semiconductor. 2024. nRF52840. https://www.nordicsemi. com/products/nrf52840

    2024

  23. [31]

    STMicroelectronics. 2024. Discovery Kit for IoT Node with STM32U5 Series. https://www.st.com/en/evaluation-tools/b-u585i-iot02a.html

    2024

  24. [32]

    STMicroelectronics. 2024. STM32U585AI. https://www.st.com/ en/microcontrollers-microprocessors/stm32u585ai.html?rt=db&id= DB4410

    2024

  25. [33]

    2017.{FarmBeats}: an{IoT} platform for{Data- Driven} agriculture

    Deepak Vasisht, Zerina Kapetanovic, Jongho Won, Xinxin Jin, Ranveer Chandra, Sudipta Sinha, Ashish Kapoor, Madhusudhan Sudarshan, and Sean Stratman. 2017.{FarmBeats}: an{IoT} platform for{Data- Driven} agriculture. In 14th USENIX Symposium on Networked Systems Design and Imple...

    2017

  26. [34]

    Veluri, C

    B. Veluri, C. Pernu, A. Saffari, J. Smith, M. Taylor, and S. Gollakota

  27. [35]

    Han Xiao, Kashif Rasul, and Roland Vollgraf. 2017. Fashion-MNIST: A Novel Image Dataset for Benchmarking Machine Learning Algorithms. arXiv preprint arXiv:1708.07747 (2017). https://arxiv.org/abs/1708. 07747 13

    2017 arXiv

  28. [2017]

    arXiv preprint arXiv:1712.05877 (2017)

    Quantizing deep convolutional networks for efficient inference: A whitepaper. arXiv preprint arXiv:1712.05877 (2017)

    2017 arXiv

  29. [2023]

    NeuriCam: Key-Frame Video Super-Resolution and Colorization for IoT Cameras. 1–17

Pith tools

Reviewed August 10, 2026 · model on record in the stance chip above.