Pith. sign in

REVIEW 3 major objections 9 minor 1 cited by

MCUCoder: Adaptive Bitrate Learned Video Compression for IoT Devices

T0 review · 3 major / 9 minor · reviewed 2026-08-12 · deepseek-v4-flash

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

Pith's one-line read The paper claims MCUCoder, a 10.5K-parameter encoder, cuts bitrate by about 55% relative to M-JPEG at equal MS-SSIM while using similar energy on IoT hardware.

desk verdict A genuinely useful IoT codec paper whose headline number is currently unverifiable because the paper never defines the MCUCoder bitstream; worth a serious referee, but the authors must fix the bitrate accounting and tone down the real-time claim. read the letter →

arxiv 2411.19442 v1 pith:TPDJEXUQ submitted 2024-11-29 eess.IV cs.CV

Ali Hojjat , Janek Haberer , Olaf Landsiedel This is my paper · ORCID
classification eess.IVcs.CV
keywords MCUCoderlearnedvideocompressionIoTcamerasadaptivebitratestreamingstochasticdropouttrainingprogressivelatentM-JPEGcomparisonINT8quantization
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

MCUCoder is a learned video compression model built for IoT cameras with only 1–2 MB of RAM. The paper tries to establish that a deliberately asymmetric design—an ultra-lightweight 10.5K-parameter INT8 encoder on the device and a larger decoder in the cloud—can deliver JPEG-level energy use while reducing bitrate by 55.65% on MCL-JCV and 55.59% on UVG, measured in MS-SSIM. It also claims that the encoder produces a 12-channel latent sorted by importance, so transmitting only the first k channels gives smooth adaptive bitrate streaming. If these claims hold, memory-constrained cameras could send roughly half the bits of M-JPEG at equal perceptual quality and still adapt to unstable networks.

What carries the argument

The central mechanism is the biased tail-drop training procedure: on each iteration a number $k$ is drawn uniformly from $(0,1)$, the last $\lfloor k \times N \rfloor$ of the $N$ latent channels are zeroed, and the decoder is trained to reconstruct from the remaining prefix. This biases the encoder to put critical information in early channels and makes the latent naturally ordered by importance. The supporting machinery is an asymmetric architecture with a three-convolution encoder, per-channel INT8 then 5-bit quantization, and a heavier decoder built from attention and residual bottleneck blocks, with INT8 quantization of weights and activations for microcontroller deployment.

What would settle it

Compress every frame of MCL-JCV and UVG on the target MCU, record the actual number of bytes transmitted for each of the 12 truncation levels, compute MS-SSIM, and compute BD-rate against M-JPEG file sizes from the same frames. If the byte count is the full 12-channel, 5-bit latent at the network's true resolution and includes no entropy coding, the 55.65% and 55.59% savings figures should reappear; otherwise the central savings claim is not reproducible.

Watch

Extended reading notes

Core claim

The paper's central discovery is that stochastic dropout training can make a tiny encoder produce an importance-ranked progressive latent. During training, a random fraction of the trailing latent channels is zeroed out before the decoder reconstructs the frame, forcing early channels to carry the most important low-frequency information and later channels to carry finer detail. At inference, the encoder quantizes each channel independently to 5-bit precision and transmits channels in order, so the bitrate-control module is only a gate that chooses how many channels to send. The authors report that the full model matches M-JPEG's energy consumption on the tested MCUs while achieving substantially better MS-SSIM per bit, and that the learned channel ordering mirrors the low-to-high frequency structure of a DCT.

Load-bearing premise

The paper never defines what counts as the MCUCoder bitstream—the latent resolution, the convolution strides, whether the 5-bit values are entropy-coded, or whether headers and packetization are included—so the headline BD-rate savings cannot be reconstructed or verified from the text alone.

Editorial extensions

If this is right

  • IoT cameras with roughly 350 KB of available RAM can run learned per-frame compression at JPEG-level energy consumption.
  • At matched MS-SSIM, the transmitted bitrate is about half of M-JPEG on the evaluated video datasets, which directly reduces bandwidth use.
  • The 12-channel progressive latent provides 12 quality levels with no added encoder computation, since the bitrate controller is just a gate.
  • The learned channel ordering behaves like a DCT-based progressive representation, which supports graceful quality degradation under packet or bandwidth loss.
  • If the reported numbers hold, adaptive bitrate learned video compression becomes practical on memory-constrained microcontrollers rather than requiring high-end hardware.

Reading between the lines

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

  • If the bitrate count is as reported, the largest end-to-end saving may come from radio transmission energy, which scales with bytes sent; the paper measures encoder energy but does not quantify this transmission saving.
  • The 12-channel choice is not fundamental to the tail-drop training, so finer adaptive granularity or even inter-frame latents are natural extensions that the paper leaves implicit.
  • The paper's bitrate appears to count the quantized latent directly; entropy-coding the 5-bit values would likely reduce the bitrate further, making the reported savings a conservative upper bound.
  • A direct way to test the ordering claim is to transmit only the first few channels over a lossy link and measure reconstruction quality; the paper's monotone MS-SSIM curves predict graceful, not catastrophic, degradation.
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.

Referee Report

3 major / 9 minor

Summary. The paper presents MCUCoder, an asymmetric learned image/video codec intended for IoT devices with 1–2 MB of RAM. The encoder is a 10.5K-parameter, INT8-quantized three-convolution network that produces 12 importance-sorted 5-bit latent channels; the decoder is a larger network intended to run on a server or cloud. The paper claims 55.65% and 55.59% BD-rate savings over M-JPEG on the MCL-JCV and UVG datasets, measured in MS-SSIM, while using approximately the same energy as M-JPEG on a nRF5340 MCU, and it reports an implementation on nRF5340 and STM32F7 with RAM/Flash measurements and 12-level adaptive bitrate truncation.

Significance. If the reported numbers hold, MCUCoder is a useful systems contribution: an extremely small encoder that runs on real MCUs, an open-source implementation, and a clean dropout-based mechanism for importance-ordered latent channels that avoids run-time content analysis. The central comparison against M-JPEG is an appropriate and meaningful external baseline, and the evaluation on real MCU hardware is a strength. The adaptive-ordering idea is inherited from the authors’ own prior ProgDTD work, so the novelty lies mainly in MCU deployment and the M-JPEG comparison rather than in the ordering mechanism itself. However, the headline bitrate claim and the real-time claim are not currently supported by the evidence presented, for the specific reasons detailed below.

major comments (3)
  1. [Section 3 (Asymmetric Compression), Figs. 7–8, Table 1] The bitrate of the MCUCoder bitstream is never defined. The manuscript does not state the spatial resolution of the 12-channel latent, the strides of the three convolutional layers, whether the transmitted values are 5-bit raw values or entropy-coded, or whether headers and packetization are included. Because BD-rate in Table 1 is a ratio of bitrates, the headline 55.65%/55.59% savings cannot be reconstructed or verified from the paper alone. Please provide the complete per-frame bitstream formula (latent height/width, strides, bit depth per value, any entropy coding, and all overhead), apply the same accounting to the M-JPEG anchor (which naturally includes JPEG headers), and recompute Figs. 7–8 and Table 1 with that definition.
  2. [Abstract, Section 4.1, Table 2] The abstract claims “smooth real-time video transmission even under fluctuating network conditions”, but Table 2 reports 237 ms per frame on the STM32F7 and 1,969 ms per frame on the nRF5340, corresponding to about 4.2 fps and 0.5 fps, respectively. The text in Section 4.1 also says the nRF5340 is “better suited for event-driven applications rather than real-time streaming”. Please remove or qualify the “real-time” claim with the achievable frame rates and the network conditions considered, and note that no dynamic bandwidth-switching experiment is presented beyond the static truncation levels in Fig. 8.
  3. [Figs. 1, 7, 8] The MS-SSIM values plotted are larger than 1 (e.g., 7.04 and 16.12 in Fig. 1), so the axis must use a transformed scale such as MS-SSIM in dB, but the paper never states this. Not only are the axis labels misleading, but Table 1’s “measured in MS-SSIM” wording is ambiguous. Please define the exact metric transformation used, label all axes and captions accordingly, and state whether the BD-rate in Table 1 is computed with this transformed metric or with the raw MS-SSIM values.
minor comments (9)
  1. [Section 3] The text says the decoder architecture is “see Fig. 12”, but Fig. 12 in Appendix C is the training-log figure; the reference should point to the correct architecture figure.
  2. [Section 4.1] The text refers to “see Table 7” for the BD-rate comparison, but no Table 7 exists; the intended reference is presumably Table 1.
  3. [Figure 1] The caption notes that images in a column do not necessarily have the same bitrate, which makes the visual comparison hard to interpret; presenting matched-bitrate pairs or annotating each image with its bpp would be more informative.
  4. [Section 4.1, Table 1] The BD-rate computation is cited only to [57]; please specify the integration interval, the number of anchor points, and the per-curve operating points used to compute Table 1.
  5. [Section 4, Fig. 7] No details are given for the JPEG/M-JPEG and H.264/H.265 encoder configurations (quality levels, rate control, chroma subsampling, preset), so the baseline curves are not reproducible; one sentence on each baseline would suffice.
  6. [Figure 9] The energy comparison reports single values (15.95 mJ vs. 14.98 mJ) with no measurement repetitions or setup description; please add this information or soften the energy claim.
  7. [Section 3, Eq. (1)] The notation N is not explicitly identified as 12 in the architecture section; it should be stated in the text rather than only appearing in the figure and equation.
  8. [Section 4] The preprocessing referred to as “noise-downsampling [55, 3]” is not described; please add a one-sentence description for reproducibility.
  9. [Table 2] The RAM/Flash numbers are presumably for the encoder only, since the decoder runs on the cloud; please state this explicitly in the table caption or surrounding text.

Circularity Check

1 steps flagged · score 2.0 of 10

Central bitrate and energy claims are externally benchmarked; only the 'importance ordering' evaluation restates its own training objective.

  1. fitted input called prediction [Section 4 ('Evaluation') and Section 4.1 ('Video compression', Figure 8 discussion)]
    "Since MCUCoder is specifically designed for IoT environments, where the structure of the output is more critical than fine details, we use MS-SSIM as the loss function. ... The results show that using more channels for decoding leads to a higher MS-SSIM, which verifies the effectiveness of the proposed stochastic dropout training."

    Stochastic dropout training optimizes reconstruction for every prefix z[0:floor(k*N)] with MS-SSIM as the loss, so the decoder is explicitly fitted to improve as more channels are provided. Thus the monotonic MS-SSIM improvement in Figure 8 is a check that training converged to the fitted objective, not an independent prediction or verification of the method. Similarly, the 'sorted by importance' property is produced by the dropout ordering during training rather than discovered independently. This self-referential evaluation does not affect the headline BD-rate savings, which are computed against the external M-JPEG/JPEG baselines.

full rationale

The paper's main quantitative claims are not circular: the 55.65%/55.59% BD-rate reductions are measured relative to M-JPEG on external datasets (MCL-JCV, UVG), the energy comparison in Figure 9 is a direct measurement on an nRF5340 against an external JPEG implementation, and the model is trained on ImageNet and evaluated on held-out datasets. The underspecified MCUCoder bitstream definition (latent resolution, strides, entropy coding, header overhead) is a serious verifiability/correctness concern, but it is an omission, not a circular reduction. The only self-referential element is the internal importance-ordering evaluation: since MS-SSIM is the training loss and the decoder is trained on randomized prefixes, showing that more channels improve MS-SSIM restates the training objective. The self-citations [50, 51] supply the dropout mechanism, but they are not used as proof of the external benchmark results, so they are not load-bearing circularity. Overall score 2 reflects this minor self-referential interpretation while the central external measurements remain independent.

Assumptions & free parameters 5 free parameters · 4 assumptions · 0 invented entities

The central claim rests on a handful of design choices (5-bit latent precision, 12 channels, uniform dropout, INT8 quantization) and domain assumptions about MCU feasibility and transfer of ImageNet training. No new physical or mathematical entities are introduced. The most fragile load-bearing premise is the unstated bitstream size calculation.

free parameters (5)
  • Latent precision = 5 bits per quantized latent value
    Chosen by hand as the bitrate-quality tradeoff for the bitstream; directly determines the reported bpp and hence the BD-rate savings.
  • Number of latent channels N = 12
    Chosen to provide 12 discrete quality levels for bandwidth adaptation; affects the granularity of adaptive streaming.
  • Dropout ratio distribution = k ~ U(0,1)
    Chosen so the encoder sees all truncation levels during training; sets the range of achievable bitrates.
  • Training resolution = 224x224
    Chosen to match IoT camera frames; all evaluation datasets are resized to this resolution.
  • Quantization scheme = INT8 weights and activations via TFLite-Micro post-training quantization
    Chosen for RAM efficiency and CMSIS-NN acceleration; the reported energy and RAM numbers depend on it.
assumptions (4)
  • domain assumption Inter-frame compression is not feasible on 1-2 MB RAM MCUs, so per-frame encoding is the only practical approach.
    Motivates the M-JPEG baseline and the per-frame design; stated in Section 1 and Section 4.1.
  • domain assumption Stochastic tail-dropout training produces a latent whose channels are ordered by importance and remain ordered at inference for unseen frames.
    The adaptive streaming claim depends on this; supported only by the paper's own training logs and Figure 8.
  • ad hoc to paper Post-training INT8 quantization to 5-bit latents does not materially change the rate-distortion tradeoff.
    The BD-rate table reports the quantized model, but no quantization-aware training is used; Figure 7 shows only small gaps, yet no error analysis is given.
  • domain assumption Training on the 300K largest ImageNet images transfers to the MCL-JCV, UVG, KODAK, and CLIC evaluation sets.
    Standard transfer assumption in learned compression; no domain shift analysis is provided.

how reviews work

0 comments
Cite this review

Pith. "Pith review of MCUCoder: Adaptive Bitrate Learned Video Compression for IoT Devices." pith.science (2026). https://pith.science/paper/TPDJEXUQ

@misc{pith2026241119442,
  author       = {Pith},
  title        = {Pith review of: MCUCoder: Adaptive Bitrate Learned Video Compression for IoT Devices},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/TPDJEXUQ}},
  note         = {Machine review of arXiv:2411.19442}
}
read the original abstract

The rapid growth of camera-based IoT devices demands the need for efficient video compression, particularly for edge applications where devices face hardware constraints, often with only 1 or 2 MB of RAM and unstable internet connections. Traditional and deep video compression methods are designed for high-end hardware, exceeding the capabilities of these constrained devices. Consequently, video compression in these scenarios is often limited to M-JPEG due to its high hardware efficiency and low complexity. This paper introduces , an open-source adaptive bitrate video compression model tailored for resource-limited IoT settings. MCUCoder features an ultra-lightweight encoder with only 10.5K parameters and a minimal 350KB memory footprint, making it well-suited for edge devices and MCUs. While MCUCoder uses a similar amount of energy as M-JPEG, it reduces bitrate by 55.65% on the MCL-JCV dataset and 55.59% on the UVG dataset, measured in MS-SSIM. Moreover, MCUCoder supports adaptive bitrate streaming by generating a latent representation that is sorted by importance, allowing transmission based on available bandwidth. This ensures smooth real-time video transmission even under fluctuating network conditions on low-resource devices. Source code available at https://github.com/ds-kiel/MCUCoder.

Figures

Figures reproduced from arXiv: 2411.19442 by the authors.

Figure 1
Figure 1. Qualitative comparison of MCUCoder and M-JPEG across various compression rates on two videos from the MCL-JCV [1] and UVG [2] datasets. As we can see, MCUCoder offers a significantly better MS-SSIM/bpp trade-off. For instance, at 0.15 bpp in the left example, with MCUCoder we can see the person’s face whereas with M-JPEG we need at least 0.34 bpp to make out the face. Note that the images in each column do not neces… view at source ↗
Figure 2
Figure 2. Number of parameters of MCUCoder and other learned image compression [3–9] and video com￾pression models [10, 11]. 1 - Limited Hardware: Although traditional video codecs like H.264 [21], H.265 [22], and the newer H.266 [23] provide excellent performance, they demand significant hardware for extracting the intra and inter-frame correlations. For exam￾ple, H.265 encoding involves highly computationally intensive task… view at source ↗
Figure 3
Figure 3. Overview of MCUCoder architecture. The encoder compresses the input frame into a sorted latent space. Afterward, channels are independently quantized and transmitted based on available bandwidth. The decoder reconstructs the frame by zeroing out missing channels. Contributions: 1. MCUCoder has an ultra-lightweight encoder with only 10.5K parameters and a minimal memory footprint of roughly 350KB RAM on nRF5340 and S… view at source ↗
Figures from the paper (10 more)
Figure 5
Figure 5. Figure 5: An example of MCUCoder bitrate adapta￾tion under dynamic network bandwidth, where the bitrate control module acts as a gate to determine the number of channels to send. NN frameworks like TFLite Micro support only a limited set of NN layers [13]. To address these const…
Figure 6
Figure 6. Figure 6: Stochas￾tic dropout training Stochastic dropout training: Bitrate adaptation is a feature that typically introduces additional complexity to the encoding process, which can be chal￾lenging to implement on MCUs due to resource constraints. In the literature, dropout [49…
Figure 7
Figure 7. Figure 7: Comparison of MCUCoder (quantized and non-quantized model) and baselines on the image (KODAK [52], CLIC [53]) and video (MCL-JCV [1], UVG [2]) compression datasets. For context, we also compare with H.264 and H.265 on video datasets, despite being impractical for MCUs …
Figure 8
Figure 8. Figure 8: MS-SSIM and bpp for the SunBath video from UVG [2] dataset. [0:k] shows the use of the first k channels (out of 12) for decoding. 5 [PITH_FULL_IMAGE:figures/full_fig_p005_8.png]
Figure 9
Figure 9. Figure 9: Energy consumption of MCUCoder compared to M-JPEG for compressing one frame on the nRF5340. Performance on MCUs: We implement MCUCoder using TFLite-Micro [56] and Zephyr RTOS [59] on STM32F7 and nRF5340 MCUs. The STM32F7 has 2 MB Flash, 2 MB RAM, and a Cortex-M7 proces…
Figure 10
Figure 10. Figure 10: Comparison of MCUCoder (quantized and non-quantized model) and baselines on the KODAK [52] and CLIC [53] datasets. All datasets are resized to 224 × 224. Since MCUCoder is specifically designed for IoT environments—where structural integrity is prioritized over fine d…
Figure 11
Figure 11. Figure 11: Comparison of MCUCoder and baselines on MCL-JCV [1] and UVG [2]. H.264 and H.265 are included for reference, though impractical for MCUs due to hardware demands. MCUCoder, optimized for IoT with a focus on MS-SSIM, prioritizes structural integrity, while M-JPEG shows …
Figure 12
Figure 12. Figure 12: MS-SSIM values on the KODAK dataset during training. The notation [PITH_FULL_IMAGE:figures/full_fig_p011_12.png]
Figure 13
Figure 13. Figure 13: Some samples from the MCL-JCV [1] dataset. The columns represent different frames, while the rows display progressively improving levels of quality from top to bottom, produced by MCUCoder. 12 [PITH_FULL_IMAGE:figures/full_fig_p012_13.png]
Figure 14
Figure 14. Figure 14: Some samples from the MCL-JCV [1] dataset. The columns represent different frames, while the rows display progressively improving levels of quality from top to bottom, produced by MCUCoder. 13 [PITH_FULL_IMAGE:figures/full_fig_p013_14.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

  1. ProCAVE: A Self-Adaptive, Full-Lifecycle Edge Caching Framework for Video Streaming via Predictive Bandwidth Estimation and Preference-Aware Deep Reinforcement Learning

    cs.NI 2026-08 conditional novelty 5.0 of 10

    A self-adaptive edge caching framework using Transformer-based bandwidth forecasting, PPO-based bitrate selection, and DDPG-based cache control improves video delivery metrics over FlyCache in simulation.

Reference graph

Works this paper leans on

60 extracted references · 50 canonical work pages · cited by 1 Pith paper

  1. [1]

    Haiqiang Wang, Weihao Gan, Sudeng Hu, Joe Yuchieh Lin, Lina Jin, Longguang Song, Ping Wang, Ioannis Katsavounidis, Anne Aaron, and C.-C. Jay Kuo. Mcl-jcv: A jnd-based h.264/avc video quality assessment dataset. In 2016 IEEE International Conference on Image Processing (ICIP), pages 1509–1513, 2016

    work page 2016

  2. [2]

    Uvg dataset: 50/120fps 4k sequences for video codec analysis and development

    Alexandre Mercat, Marko Viitanen, and Jarno Vanne. Uvg dataset: 50/120fps 4k sequences for video codec analysis and development. In Proceedings of the 11th ACM Multimedia Systems Conference, pages 297–302, 2020

    work page 2020

  3. [3]

    Variational image compression with a scale hyperprior

    Johannes Ballé, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compression with a scale hyperprior. arXiv preprint arXiv:1802.01436, 2018

    arXiv 2018

  4. [4]

    Full resolution image compression with recurrent neural networks

    George Toderici, Damien Vincent, Nick Johnston, Sung Jin Hwang, David Minnen, Joel Shor, and Michele Covell. Full resolution image compression with recurrent neural networks. In Proceedings of the IEEE conference on Computer Vision and Pattern Recognition , pages 5306–5314, 2017

    work page 2017

  5. [5]

    Dpict: Deep progressive image compression using trit-planes

    Jae-Han Lee, Seungmin Jeon, Kwang Pyo Choi, Youngo Park, and Chang-Su Kim. Dpict: Deep progressive image compression using trit-planes. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 16113–16122, 2022

    work page 2022

  6. [6]

    Context-based trit-plane coding for progressive image compression

    Seungmin Jeon, Kwang Pyo Choi, Youngo Park, and Chang-Su Kim. Context-based trit-plane coding for progressive image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14348–14357, 2023

    work page 2023

  7. [7]

    Learned image compression with mixed transformer- cnn architectures

    Jinming Liu, Heming Sun, and Jiro Katto. Learned image compression with mixed transformer- cnn architectures. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR), pages 14388–14397, June 2023

    work page 2023

  8. [8]

    Enhanced invertible encoding for learned image compression

    Yueqi Xie, Ka Leong Cheng, and Qifeng Chen. Enhanced invertible encoding for learned image compression. In Proceedings of the 29th ACM international conference on multimedia , pages 162–170, 2021

    work page 2021

Show all 60 references
  1. [9]

    Transformer-based transform coding

    Yinhao Zhu, Yang Yang, and Taco Cohen. Transformer-based transform coding. InInternational Conference on Learning Representations, 2022

    2022

  2. [10]

    Scale-space flow for end-to-end optimized video compression

    Eirikur Agustsson, David Minnen, Nick Johnston, Johannes Balle, Sung Jin Hwang, and George Toderici. Scale-space flow for end-to-end optimized video compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 8503–8512, 2020

    2020

  3. [11]

    Dvc: An end-to-end deep video compression framework

    Guo Lu, Wanli Ouyang, Dong Xu, Xiaoyun Zhang, Chunlei Cai, and Zhiyong Gao. Dvc: An end-to-end deep video compression framework. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition , pages 11006–11015, 2019

    2019

  4. [12]

    Mcunet: Tiny deep learning on iot devices

    Ji Lin, Wei-Ming Chen, Yujun Lin, Chuang Gan, Song Han, et al. Mcunet: Tiny deep learning on iot devices. Advances in neural information processing systems , 33:11711–11722, 2020

    2020

  5. [13]

    Starfish: Resilient image compression for aiot cameras

    Pan Hu, Junha Im, Zain Asgar, and Sachin Katti. Starfish: Resilient image compression for aiot cameras. In Proceedings of the 18th Conference on Embedded Networked Sensor Systems , pages 395–408, 2020

    2020

  6. [14]

    Wireless computer vision using commodity radios

    Colleen Josephson, Lei Yang, Pengyu Zhang, and Sachin Katti. Wireless computer vision using commodity radios. In Proceedings of the 18th International Conference on Information Processing in Sensor Networks, pages 229–240, 2019. 7

    2019

  7. [15]

    Towards {Battery-Free}{HD} video streaming

    Saman Naderiparizi, Mehrdad Hessar, Vamsi Talla, Shyamnath Gollakota, and Joshua R Smith. Towards {Battery-Free}{HD} video streaming. In 15th USENIX Symposium on Networked Systems Design and Implementation (NSDI 18) , pages 233–247, 2018

    2018

  8. [16]

    Neuricam: Key-frame video super-resolution and colorization for iot cameras

    Bandhav Veluri, Collin Pernu, Ali Saffari, Joshua Smith, Michael Taylor, and Shyamnath Gollakota. Neuricam: Key-frame video super-resolution and colorization for iot cameras. In Proceedings of the 29th Annual International Conference on Mobile Computing and Networking, pages 1...

    2023

  9. [17]

    Co-design of communication and machine inference for cloud robotics

    Manabu Nakanoya, Sai Shankar Narasimhan, Sharachchandra Bhat, Alexandros Anemogiannis, Akul Datta, Sachin Katti, Sandeep Chinchali, and Marco Pavone. Co-design of communication and machine inference for cloud robotics. Autonomous Robots, 47(5):579–594, 2023

    2023

  10. [18]

    Wireless steerable vision for live insects and insect-scale robots

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

    2020

  11. [19]

    Limitnet: Progressive, content- aware image offloading for extremely weak devices & networks

    Ali Hojjat, Janek Haberer, Tayyaba Zainab, and Olaf Landsiedel. Limitnet: Progressive, content- aware image offloading for extremely weak devices & networks. In Proceedings of the 22nd Annual International Conference on Mobile Systems, Applications and Services , MOBISYS ’24, ...

    2024

  12. [20]

    Wilds: A benchmark of in-the-wild distribution shifts

    Pang Wei Koh, Shiori Sagawa, Henrik Marklund, Sang Michael Xie, Marvin Zhang, Akshay Balsubramani, Weihua Hu, Michihiro Yasunaga, Richard Lanas Phillips, Irena Gao, et al. Wilds: A benchmark of in-the-wild distribution shifts. In International conference on machine learning, p...

    2021

  13. [21]

    Overview of the h

    Thomas Wiegand, Gary J Sullivan, Gisle Bjontegaard, and Ajay Luthra. Overview of the h. 264/avc video coding standard. IEEE Transactions on circuits and systems for video technology , 13(7):560–576, 2003

    2003

  14. [22]

    Overview of the high efficiency video coding (hevc) standard

    Gary J Sullivan, Jens-Rainer Ohm, Woo-Jin Han, and Thomas Wiegand. Overview of the high efficiency video coding (hevc) standard. IEEE Transactions on circuits and systems for video technology, 22(12):1649–1668, 2012

    2012

  15. [23]

    Overview of the versatile video coding (vvc) standard and its applications

    Benjamin Bross, Ye-Kui Wang, Yan Ye, Shan Liu, Jianle Chen, Gary J Sullivan, and Jens- Rainer Ohm. Overview of the versatile video coding (vvc) standard and its applications. IEEE Transactions on Circuits and Systems for Video Technology, 31(10):3736–3764, 2021

    2021

  16. [24]

    JPEG: Still image data compression standard

    William B Pennebaker and Joan L Mitchell. JPEG: Still image data compression standard . Springer Science & Business Media, 1992

    1992

  17. [25]

    Cmsis-nn: Efficient neural network kernels for arm cortex-m cpus, 2024

    ARM-software. Cmsis-nn: Efficient neural network kernels for arm cortex-m cpus, 2024. Accessed: 2024-09-22

    2024

  18. [26]

    Learning for video compres- sion with hierarchical quality and recurrent enhancement

    Ren Yang, Fabian Mentzer, Luc Van Gool, and Radu Timofte. Learning for video compres- sion with hierarchical quality and recurrent enhancement. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 6628–6637, 2020

    2020

  19. [27]

    Neural image compression via attentional multi-scale back projection and frequency decompo- sition

    Ge Gao, Pei You, Rong Pan, Shunyuan Han, Yuanyuan Zhang, Yuchao Dai, and Hojae Lee. Neural image compression via attentional multi-scale back projection and frequency decompo- sition. In Proceedings of the IEEE/CVF International Conference on Computer Vision , pages 14677–14686, 2021

    2021

  20. [28]

    Mimt: Masked image modeling transformer for video compression

    Jinxi Xiang, Kuan Tian, and Jun Zhang. Mimt: Masked image modeling transformer for video compression. In The Eleventh International Conference on Learning Representations , 2023

    2023

  21. [29]

    Vct: A video compression transformer.arXiv preprint arXiv:2206.07307, 2022

    Fabian Mentzer, George Toderici, David Minnen, Sung-Jin Hwang, Sergi Caelles, Mario Lucic, and Eirikur Agustsson. Vct: A video compression transformer.arXiv preprint arXiv:2206.07307, 2022

    2022 arXiv

  22. [30]

    A video compression framework using an overfitted restoration neural network

    Gang He, Chang Wu, Lei Li, Jinjia Zhou, Xianglin Wang, Yunfei Zheng, Bing Yu, and Weiying Xie. A video compression framework using an overfitted restoration neural network. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition Workshops, pages 14...

    2020

  23. [31]

    Instance-adaptive video compression: Improving neural codecs by training on the test set

    Ties Van Rozendaal, Johann Brehmer, Yunfan Zhang, Reza Pourreza, Auke Wiggers, and Taco S Cohen. Instance-adaptive video compression: Improving neural codecs by training on the test set. arXiv preprint arXiv:2111.10302, 2021

    2021 arXiv

  24. [32]

    Efficient video com- pression via content-adaptive super-resolution

    Mehrdad Khani, Vibhaalakshmi Sivaraman, and Mohammad Alizadeh. Efficient video com- pression via content-adaptive super-resolution. In Proceedings of the IEEE/CVF international conference on computer vision, pages 4521–4530, 2021

    2021

  25. [33]

    Coarse-to-fine deep video coding with hyperprior-guided mode prediction

    Zhihao Hu, Guo Lu, Jinyang Guo, Shan Liu, Wei Jiang, and Dong Xu. Coarse-to-fine deep video coding with hyperprior-guided mode prediction. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 5921–5930, 2022

    2022

  26. [34]

    Fvc: A new framework towards deep video compression in feature space

    Zhihao Hu, Guo Lu, and Dong Xu. Fvc: A new framework towards deep video compression in feature space. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 1502–1511, 2021

    2021

  27. [35]

    Deep contextual video compression

    Jiahao Li, Bin Li, and Yan Lu. Deep contextual video compression. Advances in Neural Information Processing Systems, 34:18114–18125, 2021

    2021

  28. [36]

    Conditional entropy coding for efficient video compression

    Jerry Liu, Shenlong Wang, Wei-Chiu Ma, Meet Shah, Rui Hu, Pranaab Dhawan, and Raquel Urtasun. Conditional entropy coding for efficient video compression. In European Conference on Computer Vision, pages 453–468. Springer, 2020

    2020

  29. [37]

    Hinerv: Video compres- sion with hierarchical encoding-based neural representation

    Ho Man Kwan, Ge Gao, Fan Zhang, Andrew Gower, and David Bull. Hinerv: Video compres- sion with hierarchical encoding-based neural representation. Advances in Neural Information Processing Systems, 36, 2024

    2024

  30. [38]

    Nerv: Neural representations for videos

    Hao Chen, Bo He, Hanyu Wang, Yixuan Ren, Ser Nam Lim, and Abhinav Shrivastava. Nerv: Neural representations for videos. Advances in Neural Information Processing Sys- tems, 34:21557–21568, 2021

    2021

  31. [39]

    A cmos image sensor and an ai accelerator for realizing edge-computing-based surveillance camera systems

    Fukashi Morishita, Norihito Kato, Satoshi Okubo, Takao Toi, Mitsuru Hiraki, Sugako Otani, Hideaki Abe, Yuji Shinohara, and Hiroyuki Kondo. A cmos image sensor and an ai accelerator for realizing edge-computing-based surveillance camera systems. In 2021 Symposium on VLSI Circui...

    2021

  32. [40]

    A 220pj/pixel/frame cmos image sensor with partial settling readout architecture

    Suyao Ji, Jing Pu, Byong Chan Lim, and Mark Horowitz. A 220pj/pixel/frame cmos image sensor with partial settling readout architecture. In 2016 IEEE Symposium on VLSI Circuits (VLSI-Circuits), pages 1–2. IEEE, 2016

    2016

  33. [41]

    Millimeter-scale ultra-low-power imaging system for intelligent edge monitoring

    Andrea Bejarano-Carbo, Hyochan An, Kyojin Choo, Shiyu Liu, Qirui Zhang, Dennis Sylvester, David Blaauw, and Hun-Seok Kim. Millimeter-scale ultra-low-power imaging system for intelligent edge monitoring. arXiv preprint arXiv:2203.04496, 2022

    2022 arXiv

  34. [42]

    7.7 a 0.2-to-3.6 tops/w programmable convolutional imager soc with in-sensor current-domain ternary-weighted mac operations for feature extraction and region-of-interest detection

    Martin Lefebvre, Ludovic Moreau, Rémi Dekimpe, and David Bol. 7.7 a 0.2-to-3.6 tops/w programmable convolutional imager soc with in-sensor current-domain ternary-weighted mac operations for feature extraction and region-of-interest detection. In 2021 IEEE International Solid-S...

    2021

  35. [43]

    4.4 a 1.3 tops/w@ 32gops fully integrated 10-core soc for iot end-nodes with 1.7µw cognitive wake-up from mram- based state-retentive sleep mode

    Davide Rossi, Francesco Conti, Manuel Eggiman, Stefan Mach, Alfio Di Mauro, Marco Guermandi, Giuseppe Tagliavini, Antonio Pullini, Igor Loi, Jie Chen, et al. 4.4 a 1.3 tops/w@ 32gops fully integrated 10-core soc for iot end-nodes with 1.7µw cognitive wake-up from mram- based s...

    2021

  36. [44]

    Macsen: A processing-in-sensor architecture integrating mac operations into image sensor for ultra-low- power bnn-based intelligent visual perception

    Han Xu, Ziru Li, Ningchao Lin, Qi Wei, Fei Qiao, Xunzhao Yin, and Huazhong Yang. Macsen: A processing-in-sensor architecture integrating mac operations into image sensor for ultra-low- power bnn-based intelligent visual perception. IEEE Transactions on Circuits and Systems II:...

    2020

  37. [45]

    Deep compressive offloading: Speeding up neural network inference by trading edge computation for network latency

    Shuochao Yao, Jinyang Li, Dongxin Liu, Tianshi Wang, Shengzhong Liu, Huajie Shao, and Tarek Abdelzaher. Deep compressive offloading: Speeding up neural network inference by trading edge computation for network latency. In Proceedings of the 18th Conference on Embedded Networke...

    2020

  38. [46]

    Elic: Efficient learned image compression with unevenly grouped space-channel contextual adaptive coding

    Dailan He, Ziming Yang, Weikun Peng, Rui Ma, Hongwei Qin, and Yan Wang. Elic: Efficient learned image compression with unevenly grouped space-channel contextual adaptive coding. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 5718–5...

    2022

  39. [47]

    Learned image compression with discretized gaussian mixture likelihoods and attention modules

    Zhengxue Cheng, Heming Sun, Masaru Takeuchi, and Jiro Katto. Learned image compression with discretized gaussian mixture likelihoods and attention modules. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages 7939–7948, 2020

    2020

  40. [48]

    Deep residual learning for image recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep residual learning for image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition , pages 770–778, 2016

    2016

  41. [49]

    Dropout: a simple way to prevent neural networks from overfitting

    Nitish Srivastava, Geoffrey Hinton, Alex Krizhevsky, Ilya Sutskever, and Ruslan Salakhutdinov. Dropout: a simple way to prevent neural networks from overfitting. The journal of machine learning research, 15(1):1929–1958, 2014

    1929

  42. [50]

    Progdtd: Progressive learned image compres- sion with double-tail-drop training

    Ali Hojjat, Janek Haberer, and Olaf Landsiedel. Progdtd: Progressive learned image compres- sion with double-tail-drop training. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) Workshops, pages 1130–1139, June 2023

    2023

  43. [51]

    Hydravit: Stacking heads for a scalable vit

    Janek Haberer, Ali Hojjat, and Olaf Landsiedel. Hydravit: Stacking heads for a scalable vit. arXiv preprint arXiv:2409.17978, 2024

    2024 arXiv

  44. [52]

    kodak lossless true color image suite (photocd pcd0992)

    Eastman kodak (1993). kodak lossless true color image suite (photocd pcd0992). https: //r0k.us/graphics/kodak

    1993

  45. [53]

    Workshop and challenge on learned image compression (clic).http://www.compression.cc,

  46. [54]

    Imagenet: A large- scale hierarchical image database

    Jia Deng, Wei Dong, Richard Socher, Li-Jia Li, Kai Li, and Li Fei-Fei. Imagenet: A large- scale hierarchical image database. In 2009 IEEE conference on computer vision and pattern recognition, pages 248–255. Ieee, 2009

    2009

  47. [55]

    Checkerboard context model for efficient learned image compression

    Dailan He, Yaoyan Zheng, Baocheng Sun, Yan Wang, and Hongwei Qin. Checkerboard context model for efficient learned image compression. In Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition, pages 14771–14780, 2021

    2021

  48. [56]

    https://www.tensorflow.org/lite/micro

    TensorFlow Lite Micro. https://www.tensorflow.org/lite/micro. Accessed: [2023]

    2023

  49. [57]

    Calculation of average psnr differences between rd-curves (vceg-m33)

    G Bjøtegaard. Calculation of average psnr differences between rd-curves (vceg-m33). In VCEG Meeting (ITU-T SG16 Q. 6), Austin, Texas, USA„ Tech. Rep. M , volume 16090, 2001

    2001

  50. [58]

    Kodak lossless true color image suite (photocd pcd0992)

    Eastman Kodak. Kodak lossless true color image suite (photocd pcd0992). http://r0k.us/ graphics/kodak, 1993. 6

    1993

  51. [59]

    https://github.com/zephyrproject-rtos/zephyr

    Zephyr Project RTOS. https://github.com/zephyrproject-rtos/zephyr. Accessed: [2023]

    2023

  52. [60]

    https://github.com/noritsuna/JPEGEncoder4Cortex-M/ tree/master

    JPEGEncoder4Cortex. https://github.com/noritsuna/JPEGEncoder4Cortex-M/ tree/master. 10 A PSNR on KODAK and CLIC datasets 0.0 0.2 0.4 0.6 0.8 Bits per Pixel (bpp) 20 22 24 26 28 30 32PSNR CLIC (224X224) JPEG MCUCoder MCUCoder(int8) 0.0 0.2 0.4 0.6 0.8 Bits per Pixel (bpp) 20 22...

Pith tools

Reviewed August 12, 2026 · model on record in the stance chip above.