Pith. sign in

REVIEW 3 major objections 4 minor 39 references

A Multi-task Supervised Compression Model for Split Computing

T0 review · 3 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 Ladon's single shared encoder serves three vision tasks at once, cutting end-to-end latency by up to 95.4% and mobile energy by up to 88.2% while matching lightweight baselines.

desk verdict A solid multi-task split-computing architecture with real hardware gains, but the unified preprocessing that underpins the single-inference claim is never specified. read the letter →

arxiv 2501.01420 v2 pith:74DCPUQ7 submitted 2025-01-02 cs.CV cs.LGeess.IV

Yoshitomo Matsubara , Matteo Mendula , Marco Levorato This is my paper · ORCID
classification cs.CVcs.LGeess.IV
keywords splitcomputingsupervisedcompressionmulti-tasklearningedgerate-distortiontradeoffimageclassificationobjectdetectionsemanticsegmentation
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

Split computing distributes a neural network across a weak mobile device and a stronger edge server: the device runs the early layers, compresses the intermediate representation, and transmits it, while the server finishes the job. The paper proposes Ladon, which it presents as the first end-to-end multi-task supervised compression model for that setting. Ladon shares one encoder and one image preprocessing pipeline across image classification, object detection, and semantic segmentation, so a single inference produces all three outputs. The authors report that Ladon matches or outperforms lightweight single-task baselines on ILSVRC 2012, COCO 2017, and PASCAL VOC 2012, while cutting end-to-end latency by up to 95.4% and mobile-device energy by up to 88.2% compared with baselines. If correct, this removes a barrier to running several vision tasks simultaneously on battery-powered, bandwidth-limited edge systems.

What carries the argument

Supervised compression—learning a compressed representation that carries task-relevant information rather than reconstructing the input—is implemented here by an early-layer encoder $f_\theta$ that maps an input image to a quantized representation $\hat{z}$, and a decoder-plus-heads module $g_\phi$ that maps $\hat{z}$ to predictions. Training minimizes a rate-distortion loss, $$L_{\mathrm{pre}}(x) = \sum_{i \in I} \|h^t_i(x) - h^s_i(x)\|$_2^{2}$ - \$\beta$ \log p_\psi(f_\$\theta$(x) + \epsilon),$$ where the first term distills teacher embeddings into student layers, the second term charges for the bit cost of the quantized representation via an entropy model, and $\epsilon \sim \mathrm{Unif}(-1/2, 1/2)$ simulates quantization noise. The structural novelty is that one encoder, one shared backbone, and one unified preprocessing pipeline feed three heads—a classification head, a Faster R-CNN with FPN detection head, and a DeepLabv3 segmentation head—trained in three steps: pretrain the encoder-decoder by distillation, fine-tune for classification, then freeze and attach the extra heads.

What would settle it

Run Ladon's detection and segmentation heads on COCO 2017 and PASCAL VOC 2012 under the paper's unified preprocessing, then rerun the same heads under each task's conventional preprocessing (full image, aspect ratio preserved, no center crop). If the unified-pipeline versions drop by more than a point or two of mAP or mIoU, especially on small objects and boundary pixels, the paper's claim that the unified pipeline costs no accuracy is falsified. A simpler probe: feed Ladon's detection head a standard 224×224 center-cropped input and compare with the same head on full-resolution input; any large drop isolates the preprocessing assumption rather than the shared encoder.

Watch

Extended reading notes

Core claim

On the authors' own terms, the central claim is that supervised compression need not be trained per task: a single encoder trained to emit small, task-relevant representations at the early layers can feed shared backbone parameters and three task-specific heads, eliminating the duplicate local computation, duplicate transmission, and separate preprocessing of prior split-computing approaches. Concretely, with ResNet-50 and ResNeSt-269e backbones and a rate-distortion loss, Ladon's encoders are only 0.543--0.935 MB, roughly 0.268--2.29% of the total multi-task local model size, and the paper reports accuracy on ILSVRC 2012 and COCO 2017 that is competitive with or better than mobile-friendly baselines, with semantic segmentation comparable. On low-power mobile devices with a 100 Kbps or LoRa-rate wireless link, the measured end-to-end latency falls by up to 95.4% and mobile-device energy by 65.0--88.2% relative to baselines. The paper's conclusion is that a deliberate multi-task supervised compression design, not a collection of task-specific models sharing a generic encoder, is the efficient way to serve multi-task inference in split computing.

Load-bearing premise

The load-bearing premise is that a single image preprocessing pipeline—one resizing and cropping choice—can serve image classification, object detection, and semantic segmentation without meaningful accuracy loss; the paper asserts this but never specifies the pipeline, and the three tasks normally favor different preprocessing.

Editorial extensions

If this is right

  • A multi-task query costs about the same communication as a single-task query: one compressed representation $\hat{z}$ is transmitted once and decoded into classification labels, bounding boxes, and segmentations.
  • Edge-server memory and compute shrink because one shared backbone replaces three independently trained backbones, letting a fixed server host more tasks or more concurrent users.
  • On mobile devices, the energy savings of 65.0--88.2% directly extend battery life, and the paper's LoRa-rate experiments show the latency advantage persists even at 37.5 Kbps.
  • The hyperparameter $\beta$ in the rate-distortion loss becomes a system-level knob, letting operators trade compressed data size against task accuracy while keeping the encoder small.
  • The three-step training recipe—pretrain with distillation, fine-tune for classification, then freeze and attach extra heads—is a reusable path for adding new task heads to an existing compressed encoder.

Reading between the lines

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

  • The same one-encoder recipe could plausibly extend to related dense prediction heads such as depth estimation or panoptic segmentation, since those tasks also consume shared backbone features; the paper itself only demonstrates classification, detection, and segmentation.
  • Because semantic segmentation is the weakest of the three reported results, the number of tasks one encoder can serve is likely bounded by the most resolution-sensitive task in the mix, not by the average task.
  • A component-wise ablation—shared encoder only versus shared encoder plus unified preprocessing—would separate how much of the 95.4% latency saving comes from avoiding duplicate backbones and how much from avoiding duplicate preprocessing.
  • The unified preprocessing pipeline is underspecified in the paper; a reader reproducing the work should first pin down the exact resizing and cropping rule, since it determines whether the accuracy-parity claim holds on small objects.
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 / 4 minor

Summary. The paper proposes Ladon, a supervised compression model for split computing that shares a single encoder and a unified preprocessing pipeline across three vision tasks: image classification (ILSVRC 2012), object detection (COCO 2017), and semantic segmentation (PASCAL VOC 2012). The encoder is executed on a mobile device and transmits a compressed representation to an edge server, which runs task-specific heads. The authors evaluate predictive performance against lightweight single-task baselines, and measure end-to-end latency and mobile energy consumption on NVIDIA Jetson devices, reporting large reductions relative to local computing and to a split-computing baseline (Entropic Student). The paper claims to present the first end-to-end multi-task supervised compression model for split computing.

Significance. If the central claims hold, Ladon would be a useful contribution: a single encoded transmission serving three tasks in one inference, with publicly available code and models, and real-device measurements of latency and energy. The use of the SC2 benchmark and the inclusion of Jetson measurements are strengths. However, several load-bearing points are underspecified or unsupported as written, notably the unified preprocessing pipeline, the 'end-to-end' training description, and the baseline selection for isolating the effect of the proposed design choices.

major comments (3)
  1. [Section 4.1 (Unified preprocessing)] The unified preprocessing pipeline is never specified. The text describes typical task-specific preprocessing (square center-crop for classification, aspect-preserving resizing for detection and segmentation) but does not state which policy Ladon uses: the unified input resolution, the resize/crop rules, how aspect ratio is handled, or how detection/segmentation outputs are mapped back to the original image coordinates. The claim that this unified pipeline works 'without model accuracy degradation' is also not tested against a control: the only predictive performance comparisons in Table 2 are against different single-task lightweight baselines with task-specific preprocessing. To support the central claim that one encoded transmission can serve all tasks, the authors should specify the pipeline and report an ablation comparing Ladon with the unified pipeline against Ladon with task-specific preprocessing for each head.
  2. [Section 4.3 (Training) / Introduction] The paper calls Ladon an 'end-to-end multi-task supervised compression model' (Introduction, Section 1) and an 'end-to-end multi-task model' (Figure 1), but the training procedure in Steps 1-3 is sequential: encoder-decoder pre-training, then classification fine-tuning, then freezing all parameters and separately fine-tuning the detection and segmentation heads. There is no joint multi-task loss and no simultaneous optimization of the shared representation for the three tasks. The phrase 'end-to-end multi-task' is therefore not supported by the described training protocol. Please either revise the claim to 'multi-task-head supervised compression model' or provide a genuinely joint training scheme and demonstrate its impact.
  3. [Section 5.2 / Table 2 / Figures 2-4] The predictive performance evaluation uses only lightweight single-task baselines (MobileNetV2/MobileNetV3, MNASNet, SSD, DeepLab, LRASSP) as references, as summarized in Table 2. This does not isolate the effects of the two central design choices: the unified preprocessing pipeline and the shared multi-task backbone. To support the claim that the unified pipeline does not degrade accuracy and that parameter sharing does not hurt task-specific performance, the paper should add ablations comparing Ladon against: (a) Ladon with task-specific preprocessing for each head, (b) the same backbone (ResNet-50 or ResNeSt-269e) trained as independent single-task SC2 models without sharing, and (c) full-offloading execution with the same backbones. Without these controls, the reported 'matching or outperforming' results cannot be attributed to the proposed architecture and preprocessing choices.
minor comments (4)
  1. [Section 6.2 / 6.3] The latency and energy experiments report point estimates without confidence intervals or number of repeated runs. Given that the central quantitative claims are reductions of up to 95.4% in latency and 88.2% in energy, adding error bars or standard deviations would strengthen the reliability of these measurements.
  2. [Section 4.3 / Supplementary A] The paper uses the notation gϕ both for the 'remaining layers' that produce a prediction and for the decoder in Equation (1). Clarifying this notation would avoid ambiguity about whether the decoder is part of the encoder-side or the task-specific modules.
  3. [Section 6.1] The claim that 'Ladon models either outperformed or rivaled' the baselines is supported by Figures 2-4, but the semantic segmentation results are described as 'comparable' rather than superior. Please state explicitly in the text which Ladon configurations (backbone and β) achieve which qualitative conclusion on each dataset, since the figures alone do not convey the exact performance ranges.
  4. [Section 3 / Introduction] The comparison with Chimera is described only in terms of its bottleneck assumption. Since Ladon is claimed to be the 'first' end-to-end multi-task supervised compression model, consider adding a sentence explaining explicitly why Chimera (which also serves multiple tasks with a shared feature extractor) does not fall under this definition, beyond the bottleneck criterion.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: Ladon's accuracy and latency claims are empirical evaluations against external baselines, not reductions to fitted inputs or self-citation chains.

full rationale

The paper's central claims are evaluated empirically: the Ladon encoder is pretrained with a standard rate-distortion loss, fine-tuned for classification, and then task-specific detection and segmentation heads are trained on COCO and VOC with the encoder frozen. Predictive performance is benchmarked against independent lightweight models (MobileNetV3, SSD, DeepLabv3, etc.) on external datasets, and latency and energy are measured on real Jetson devices. The beta sweep produces a rate-distortion curve rather than a fitted quantity that is later relabeled as a prediction; no equation in the paper defines the reported accuracy in terms of the model's own hyperparameters. The main weakness—the unified preprocessing pipeline is asserted to incur 'no model accuracy degradation' without specifying the pipeline or ablating it against task-specific preprocessing—is an underspecification and correctness concern, not circularity, because the claim is not true by construction and is in principle falsifiable from the released code. Self-citations to prior supervised compression work supply the encoder architecture and benchmark protocol; both are code-released and externally reproducible, so they do not function as unverified premises that force the conclusion.

Assumptions & free parameters 1 free parameters · 3 assumptions · 0 invented entities

The central claim rests on transferring a classification-trained compressed representation to detection and segmentation, and on the viability of a unified preprocessing pipeline. No new physical entities or concepts are introduced. The main free parameter is beta, which traces a tradeoff curve.

free parameters (1)
  • Rate-distortion tradeoff coefficient beta (per architecture) = 0.32, 0.64, 1.28, 2.56, 5.12 (ResNet-50 and ResNeSt-269e variants)
    Hyperparameter controlling compressed data size versus distortion in Eq. (1). Multiple values are reported to draw a tradeoff curve, not a single fitted constant. The central claim does not hinge on one value.
assumptions (3)
  • domain assumption The shared encoder trained for image classification learns compressed representations at early layers that are informative enough for object detection and semantic segmentation after head fine-tuning.
    In Step 3 the encoder and backbone are frozen and only task-specific heads are updated. If classification-trained features do not transfer to dense prediction tasks, the multi-task accuracy claim fails.
  • domain assumption A single unified image preprocessing pipeline can serve classification, detection, and segmentation without significant accuracy loss.
    The paper asserts this 'without model accuracy degradation' but never specifies the pipeline. Tasks normally use different preprocessing (center-crop vs. aspect-ratio-preserving resizing), so this is load-bearing.
  • domain assumption The rate-distortion loss with uniform quantization noise from Eq. (1), inherited from the entropic student, is an effective learning objective for the multi-task setting.
    The loss is taken from prior work [23, 24]; the paper does not re-derive or validate it for multiple heads.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A Multi-task Supervised Compression Model for Split Computing." pith.science (2026). https://pith.science/paper/74DCPUQ7

@misc{pith2026250101420,
  author       = {Pith},
  title        = {Pith review of: A Multi-task Supervised Compression Model for Split Computing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/74DCPUQ7}},
  note         = {Machine review of arXiv:2501.01420}
}
abstract

Split computing ($\neq$ split learning) is a promising approach to deep learning models for resource-constrained edge computing systems, where weak sensor (mobile) devices are wirelessly connected to stronger edge servers through channels with limited communication capacity. State-of-theart work on split computing presents methods for single tasks such as image classification, object detection, or semantic segmentation. The application of existing methods to multitask problems degrades model accuracy and/or significantly increase runtime latency. In this study, we propose Ladon, the first multi-task-head supervised compression model for multi-task split computing. Experimental results show that the multi-task supervised compression model either outperformed or rivaled strong lightweight baseline models in terms of predictive performance for ILSVRC 2012, COCO 2017, and PASCAL VOC 2012 datasets while learning compressed representations at its early layers. Furthermore, our models reduced end-to-end latency (by up to 95.4%) and energy consumption of mobile devices (by up to 88.2%) in multi-task split computing scenarios.

Figures

Figures reproduced from arXiv: 2501.01420 by the authors.

Figure 1
Figure 1. Entropic student (top) vs. our Ladon (bottom) in multi [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. ILSVRC 2012: tradeoff between compressed data size [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. COCO 2017: tradeoff between compressed data size and [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: PASCAL VOC 2012: tradeoff between compressed data [PITH_FULL_IMAGE:figures/full_fig_p005_4.png]
Figure 5
Figure 5. Figure 5: End-to-end latency for Jetson Nano (mobile device) and laptop with CUDA (edge server). Top/bottom: local computing [PITH_FULL_IMAGE:figures/full_fig_p007_5.png]
Figure 6
Figure 6. Figure 6: End-to-end latency for Jetson NX Xavier (mobile device) and laptop with CUDA (edge server). Top/bottom: local computing [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Energy consumption of Jetson Nano (mobile device). Top/bottom: local computing without/with CUDA. [PITH_FULL_IMAGE:figures/full_fig_p008_7.png]
Figure 8
Figure 8. Figure 8: Energy consumption of Jetson NX Xavier (mobile device). Top/bottom: local computing without/with CUDA. [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

39 extracted references · 37 canonical work pages

  1. [1]

    EneA-FL: Energy-aware orchestration for serverless federated learning

    Andrea Agiollo, Paolo Bellavista, Matteo Mendula, and An- drea Omicini. EneA-FL: Energy-aware orchestration for serverless federated learning. Future Generation Computer Systems, 154:219–234, 2024. 5

    work page 2024

  2. [2]

    End- to-end Optimized Image Compression

    Johannes Ball´e, Valero Laparra, and Eero P Simoncelli. End- to-end Optimized Image Compression. International Confer- ence on Learning Representations, 2017. 2

    work page 2017

  3. [3]

    Variational image compression with a scale hyperprior

    Johannes Ball ´e, David Minnen, Saurabh Singh, Sung Jin Hwang, and Nick Johnston. Variational image compression with a scale hyperprior. InInternational Conference on Learn- ing Representations, 2018. 2, 4

    work page 2018

  4. [4]

    Rethinking Atrous Convolution for Semantic Image Segmentation

    Liang-Chieh Chen, George Papandreou, Florian Schroff, and Hartwig Adam. Rethinking Atrous Convolution for Semantic Image Segmentation. arXiv preprint arXiv:1706.05587, 2017. 3, 4, 11

    arXiv 2017

  5. [5]

    A Performance Model for GPUs with Caches

    Thanh Tuan Dao, Jungwon Kim, Sangmin Seo, Bernhard Egger, and Jaejin Lee. A Performance Model for GPUs with Caches. IEEE Transactions on Parallel and Distributed Systems, 26(7):1800–1813, 2014. 5

    work page 2014

  6. [6]

    Design and Ex- perimental Validation of a Distributed Interaction Protocol for Connected Autonomous Vehicles at a Road Intersection

    Marco Di Vaio, Paolo Falcone, Robert Hult, Alberto Petrillo, Alessandro Salvi, and Stefania Santini. Design and Ex- perimental Validation of a Distributed Interaction Protocol for Connected Autonomous Vehicles at a Road Intersection. IEEE Transactions on Vehicular Technology, 68(10):9451– 9465, 2019. 1

    work page 2019

  7. [7]

    BottleNet: A Deep Learning Architecture for Intelligent Mobile Cloud Computing Services

    Amir Erfan Eshratifar, Amirhossein Esmaili, and Massoud Pedram. BottleNet: A Deep Learning Architecture for Intelligent Mobile Cloud Computing Services. In 2019 IEEE/ACM Int. Symposium on Low Power Electronics and Design (ISLPED), pages 1–6, 2019. 2

    work page 2019

  8. [8]

    The PASCAL Visual Object Classes Challenge 2012 (VOC2012)

    Mark Everingham, Luc Van Gool, CKI Williams, John Winn, and Andrew Zisserman. The PASCAL Visual Object Classes Challenge 2012 (VOC2012). 2012. 1, 2, 4, 11

    work page 2012

Show all 39 references
  1. [9]

    Rethinking ImageNet Pre-training

    Kaiming He, Ross Girshick, and Piotr Doll ´ar. Rethinking ImageNet Pre-training. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 4918– 4927, 2019. 4

    2019

  2. [10]

    Deep Residual Learning for Image Recognition

    Kaiming He, Xiangyu Zhang, Shaoqing Ren, and Jian Sun. Deep Residual Learning for Image Recognition. In Proceed- ings of the IEEE Conference on Computer Vision and Pattern Recognition, pages 770–778, 2016. 3

    2016

  3. [11]

    Distilling the Knowledge in a Neural Network

    Geoffrey Hinton, Oriol Vinyals, and Jeff Dean. Distilling the Knowledge in a Neural Network. In Deep Learning and Representation Learning Workshop: NIPS 2014, 2014. 4, 11

    2014

  4. [12]

    Searching for Mo- bileNetV3

    Andrew Howard, Mark Sandler, Grace Chu, Liang-Chieh Chen, Bo Chen, Mingxing Tan, Weijun Wang, Yukun Zhu, Ruoming Pang, Vijay Vasudevan, et al. Searching for Mo- bileNetV3. In Proceedings of the IEEE/CVF International Conference on Computer Vision, pages 1314–1324, 2019. 4

    2019

  5. [13]

    Kingma and Jimmy Ba

    Diederik P. Kingma and Jimmy Ba. Adam: A Method for Stochastic Optimization. In Third International Conference on Learning Representations, 2015. 11

    2015

  6. [14]

    Auto-Encoding Varia- tional Bayes

    Diederik P Kingma and Max Welling. Auto-Encoding Varia- tional Bayes. In International Conference on Learning Rep- resentations, 2014. 1

    2014

  7. [15]

    Feature Pyramid Networks for Object Detection

    Tsung-Yi Lin, Piotr Doll ´ar, Ross Girshick, Kaiming He, Bharath Hariharan, and Serge Belongie. Feature Pyramid Networks for Object Detection. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition , pages 2117–2125, 2017. 3, 11

    2017

  8. [16]

    Microsoft COCO: Common Objects in Context

    Tsung-Yi Lin, Michael Maire, Serge Belongie, James Hays, Pietro Perona, Deva Ramanan, Piotr Doll´ar, and C Lawrence Zitnick. Microsoft COCO: Common Objects in Context. In European conference on computer vision , pages 740–755. Springer, 2014. 1, 2, 4, 11

    2014

  9. [17]

    SSD: Single shot multibox detector

    Wei Liu, Dragomir Anguelov, Dumitru Erhan, Christian Szegedy, Scott Reed, Cheng-Yang Fu, and Alexander C Berg. SSD: Single shot multibox detector. In European conference on computer vision, pages 21–37, 2016. 4

    2016

  10. [18]

    Deep Learning Face Attributes in the Wild

    Ziwei Liu, Ping Luo, Xiaogang Wang, and Xiaoou Tang. Deep Learning Face Attributes in the Wild. In Proceedings of the IEEE international conference on computer vision, pages 3730–3738, 2015. 2

    2015

  11. [19]

    Distilled Split Deep Neural Networks for Edge-Assisted Real-Time Systems

    Yoshitomo Matsubara, Sabur Baidya, Davide Callegaro, Marco Levorato, and Sameer Singh. Distilled Split Deep Neural Networks for Edge-Assisted Real-Time Systems. In Proceedings of the 2019 Workshop on Hot Topics in Video Analytics and Intelligent Edges, pages 21–26, 2019. 2

    2019

  12. [20]

    Head Network Dis- tillation: Splitting Distilled Deep Neural Networks for Resource-Constrained Edge Computing Systems

    Yoshitomo Matsubara, Davide Callegaro, Sabur Baidya, Marco Levorato, and Sameer Singh. Head Network Dis- tillation: Splitting Distilled Deep Neural Networks for Resource-Constrained Edge Computing Systems. IEEE Ac- cess, 8:212177–212193, 2020. 1

    2020

  13. [21]

    Neural Com- pression and Filtering for Edge-assisted Real-time Object De- tection in Challenged Networks

    Yoshitomo Matsubara and Marco Levorato. Neural Com- pression and Filtering for Edge-assisted Real-time Object De- tection in Challenged Networks. In 2020 25th International Conference on Pattern Recognition (ICPR), pages 2272–2279,

    2020

  14. [22]

    Split Computing and Early Exiting for Deep Learn- ing Applications: Survey and Research Challenges

    Yoshitomo Matsubara, Marco Levorato, and Francesco Restuccia. Split Computing and Early Exiting for Deep Learn- ing Applications: Survey and Research Challenges. ACM Computing Surveys, 55(5):1–30, 2022. 1, 2

    2022

  15. [23]

    Supervised Compression for Resource- Constrained Edge Computing Systems

    Yoshitomo Matsubara, Ruihan Yang, Marco Levorato, and Stephan Mandt. Supervised Compression for Resource- Constrained Edge Computing Systems. In Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision, pages 2685–2695, 2022. 2, 3, 5

    2022

  16. [24]

    SC2 Benchmark: Supervised Compression for Split Computing

    Yoshitomo Matsubara, Ruihan Yang, Marco Levorato, and Stephan Mandt. SC2 Benchmark: Supervised Compression for Split Computing. Transactions on Machine Learning Research, 2023. 1, 2, 4, 5, 6

    2023

  17. [25]

    Communication- Efficient Learning of Deep Networks from Decentralized Data

    Brendan McMahan, Eider Moore, Daniel Ramage, Seth Hampson, and Blaise Aguera y Arcas. Communication- Efficient Learning of Deep Networks from Decentralized Data. In Artificial intelligence and statistics , pages 1273–

  18. [26]

    Furcifer: a Context Adaptive Middleware for Real-world Object Detection Ex- ploiting Local, Edge, and Split Computing in the Cloud Con- tinuum

    Matteo Mendula, Paolo Bellavista, Marco Levorato, and Sharon Ladron de Guevara Contreras. Furcifer: a Context Adaptive Middleware for Real-world Object Detection Ex- ploiting Local, Edge, and Split Computing in the Cloud Con- tinuum. In 2024 IEEE International Conference on Pe...

    2024

  19. [27]

    Chimera: Context-Aware Splittable Deep Multitasking Models for Edge Intelligence

    Sumaiya Tabassum Nimi, Md Adnan Arefeen, Md Yusuf Sar- war Uddin, Biplob Debnath, and Srimat Chakradhar. Chimera: Context-Aware Splittable Deep Multitasking Models for Edge Intelligence. In 2022 IEEE International Conference on Smart Computing (SMARTCOMP), pages 70–77. IEEE, 2022. 2

    2022

  20. [28]

    PyTorch: An imperative style, high-performance deep learning library

    Adam Paszke, Sam Gross, Francisco Massa, Adam Lerer, James Bradbury, Gregory Chanan, Trevor Killeen, Zeming Lin, Natalia Gimelshein, Luca Antiga, et al. PyTorch: An imperative style, high-performance deep learning library. In Advances in Neural Information Processing Systems, ...

    2019

  21. [29]

    Prometheus: A next- generation monitoring system (talk)

    Bjorn Rabenstein and Julius V olz. Prometheus: A next- generation monitoring system (talk). Dublin, May 2015. USENIX Association. 5

    2015

  22. [30]

    Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks

    Shaoqing Ren, Kaiming He, Ross Girshick, and Jian Sun. Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networks. In Advances in Neural Informa- tion Processing Systems, pages 91–99, 2015. 3, 4, 11

    2015

  23. [31]

    Berg, and Li Fei-Fei

    Olga Russakovsky, Jia Deng, Hao Su, Jonathan Krause, San- jeev Satheesh, Sean Ma, Zhiheng Huang, Andrej Karpathy, Aditya Khosla, Michael Bernstein, Alexander C. Berg, and Li Fei-Fei. ImageNet Large Scale Visual Recognition Challenge. International Journal of Computer Vision, 1...

  24. [32]

    IoT Technologies for Embedded Computing: A Survey

    Farzad Samie, Lars Bauer, and J¨org Henkel. IoT Technologies for Embedded Computing: A Survey. In 2016 International Conference on Hardware/Software Codesign and System Syn- thesis (CODES+ ISSS), pages 1–10. IEEE, 2016. 4, 11

    2016

  25. [33]

    MobileNetV2: Inverted Residuals and Linear Bottlenecks

    Mark Sandler, Andrew Howard, Menglong Zhu, Andrey Zh- moginov, and Liang-Chieh Chen. MobileNetV2: Inverted Residuals and Linear Bottlenecks. In Proceedings of the IEEE Conference on Computer Vision and Pattern Recogni- tion, pages 4510–4520, 2018. 4

    2018

  26. [34]

    Indoor Segmentation and Support Inference from RGBD Images

    Nathan Silberman, Derek Hoiem, Pushmeet Kohli, and Rob Fergus. Indoor Segmentation and Support Inference from RGBD Images. In Computer Vision – ECCV 2012 , pages 746–760. Springer, 2012. 2

    2012

  27. [35]

    End-to- end Learning of Compressible Features

    Saurabh Singh, Sami Abu-El-Haija, Nick Johnston, Johannes Ball´e, Abhinav Shrivastava, and George Toderici. End-to- end Learning of Compressible Features. In 2020 IEEE In- ternational Conference on Image Processing (ICIP), pages 3349–3353. IEEE, 2020. 2

    2020

  28. [36]

    Area under a Curve: Trapezoidal and Simpson’s Rules

    Ronald J Tallarida, Rodney B Murray, Ronald J Tallarida, and Rodney B Murray. Area under a Curve: Trapezoidal and Simpson’s Rules. Manual of Pharmacologic Calculations: with Computer Programs, pages 77–81, 1987. 5

    1987

  29. [37]

    MnasNet: Platform-Aware Neural Architecture Search for Mobile

    Mingxing Tan, Bo Chen, Ruoming Pang, Vijay Vasudevan, Mark Sandler, Andrew Howard, and Quoc V Le. MnasNet: Platform-Aware Neural Architecture Search for Mobile. In Proceedings of the IEEE Conf. on Computer Vision and Pat- tern Recognition, pages 2820–2828, 2019. 4

    2019

  30. [38]

    Split learning for health: Distributed deep learning without sharing raw patient data

    Praneeth Vepakomma, Otkrist Gupta, Tristan Swedish, and Ramesh Raskar. Split learning for health: Distributed deep learning without sharing raw patient data. arXiv preprint arXiv:1812.00564, 2018. 1, 2

    2018 arXiv

  31. [39]

    ResNeSt: Split-Attention Networks

    Hang Zhang, Chongruo Wu, Zhongyue Zhang, Yi Zhu, Haibin Lin, Zhi Zhang, Yue Sun, Tong He, Jonas Mueller, R Man- matha, et al. ResNeSt: Split-Attention Networks. In Proceed- ings of the IEEE/CVF conference on computer vision and pattern recognition, pages 2736–2746, 2022. 3, 11...

    2022

Pith tools

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