{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,3,27]],"date-time":"2026-03-27T09:13:18Z","timestamp":1774602798281,"version":"3.50.1"},"reference-count":47,"publisher":"Springer Science and Business Media LLC","issue":"3","license":[{"start":{"date-parts":[[2026,1,29]],"date-time":"2026-01-29T00:00:00Z","timestamp":1769644800000},"content-version":"tdm","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"},{"start":{"date-parts":[[2026,1,29]],"date-time":"2026-01-29T00:00:00Z","timestamp":1769644800000},"content-version":"vor","delay-in-days":0,"URL":"https:\/\/creativecommons.org\/licenses\/by\/4.0"}],"funder":[{"DOI":"10.13039\/501100005073","name":"Agency for Defense Development","doi-asserted-by":"publisher","award":["915062201"],"award-info":[{"award-number":["915062201"]}],"id":[{"id":"10.13039\/501100005073","id-type":"DOI","asserted-by":"publisher"}]},{"DOI":"10.13039\/501100010418","name":"Institute for Information and Communications Technology Promotio","doi-asserted-by":"publisher","award":["RS-2024-00508465"],"award-info":[{"award-number":["RS-2024-00508465"]}],"id":[{"id":"10.13039\/501100010418","id-type":"DOI","asserted-by":"publisher"}]}],"content-domain":{"domain":["link.springer.com"],"crossmark-restriction":false},"short-container-title":["Int J Comput Vis"],"published-print":{"date-parts":[[2026,3]]},"abstract":"<jats:title>Abstract<\/jats:title>\n                  <jats:p>\n                    In this paper, we introduce a memory-efficient CNN (convolutional neural network), which enables resource-constrained low-end embedded and IoT devices to perform on-device vision and audio tasks, such as image classification, object detection, and audio classification, using extremely low memory,\n                    <jats:italic>i.e<\/jats:italic>\n                    ., only 63 KB on ImageNet classification. Based on the bottleneck block of MobileNet, we propose three design principles that significantly curtail the peak memory usage of a CNN so that it can fit the limited KB memory of the low-end device. First, \u2018input segmentation\u2019 divides an input image into a set of patches, including the central patch overlapped with the others, reducing the size (and memory requirement) of a large input image. Second, \u2018patch tunneling\u2019 builds independent tunnel-like paths consisting of multiple bottleneck blocks per patch, penetrating through the entire model from an input patch to the last layer of the network, maintaining lightweight memory usage throughout the whole network. Lastly, \u2018bottleneck reordering\u2019 rearranges the execution order of convolution operations inside the bottleneck block such that the memory usage remains constant regardless of the size of the convolution output channels. We also present \u2018peak memory aware quantization\u2019, enabling desired peak memory reduction in actual deployment of quantized network. The experiment result shows that the proposed network classifies ImageNet with extremely low memory (\n                    <jats:italic>i.e<\/jats:italic>\n                    ., 63 KB) while achieving competitive top-1 accuracy (\n                    <jats:italic>i.e<\/jats:italic>\n                    ., 61.58%). To the best of our knowledge, the memory usage of the proposed network is far smaller than state-of-the-art memory-efficient networks,\n                    <jats:italic>i.e<\/jats:italic>\n                    ., up to 89x and 3.1x smaller than MobileNet (\n                    <jats:italic>i.e<\/jats:italic>\n                    ., 5.6 MB) and MCUNet (\n                    <jats:italic>i.e<\/jats:italic>\n                    ., 196 KB), respectively.\n                  <\/jats:p>","DOI":"10.1007\/s11263-025-02688-w","type":"journal-article","created":{"date-parts":[[2026,1,29]],"date-time":"2026-01-29T08:48:59Z","timestamp":1769676539000},"update-policy":"https:\/\/doi.org\/10.1007\/springer_crossmark_policy","source":"Crossref","is-referenced-by-count":0,"title":["Designing Extremely Memory-Efficient CNNs for On-device Vision and Audio Tasks"],"prefix":"10.1007","volume":"134","author":[{"given":"Yoel","family":"Park","sequence":"first","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"given":"Jaewook","family":"Lee","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]},{"ORCID":"https:\/\/orcid.org\/0009-0004-7162-0845","authenticated-orcid":false,"given":"Seulki","family":"Lee","sequence":"additional","affiliation":[],"role":[{"role":"author","vocabulary":"crossref"}]}],"member":"297","published-online":{"date-parts":[[2026,1,29]]},"reference":[{"key":"2688_CR1","unstructured":"Arm Holdings plc (2014). ARM Coretex-M7. https:\/\/www.arm.com\/products\/silicon-ip-cpu\/cortex-m\/cortex-m7."},{"issue":"5","key":"2688_CR2","doi-asserted-by":"publisher","first-page":"433","DOI":"10.1016\/j.mejo.2005.07.002","volume":"37","author":"M Bigas","year":"2006","unstructured":"Bigas, M., Cabruja, E., Forest, J., & Salvi, J. (2006). Review of cmos image sensors. Microelectronics journal,37(5), 433\u2013451.","journal-title":"Microelectronics journal"},{"key":"2688_CR3","doi-asserted-by":"crossref","unstructured":"Chaturvedi, A., Yadav, S. A., Salman, H. M., Goyal, H. R., Gebregziabher, H., & Rao, A. K. (2022). Classification of sound using convolutional neural networks. (pp. 1015\u20131019)","DOI":"10.1109\/IC3I56241.2022.10072823"},{"issue":"1","key":"2688_CR4","doi-asserted-by":"publisher","first-page":"130","DOI":"10.1109\/JSSC.2015.2470526","volume":"51","author":"J Choi","year":"2015","unstructured":"Choi, J., Shin, J., Kang, D., & Park, D.-S. (2015). Always-on cmos image sensor for mobile and wearable devices. IEEE Journal of Solid-State Circuits,51(1), 130\u2013140.","journal-title":"IEEE Journal of Solid-State Circuits"},{"key":"2688_CR5","unstructured":"Chowdhery, A., Warden, P., Shlens, J., Howard, A., & Rhodes, R. (2019). Visual wake words dataset."},{"key":"2688_CR6","doi-asserted-by":"crossref","unstructured":"Cordts, M., Omran, M., Ramos, S., Rehfeld, T., Enzweiler, M., Benenson, R., Franke, U., Roth, S., & Schiele, B. (2016). The cityscapes dataset for semantic urban scene understanding.","DOI":"10.1109\/CVPR.2016.350"},{"key":"2688_CR7","first-page":"248","volume-title":"Imagenet: A large-scale hierarchical image database","author":"J Deng","year":"2009","unstructured":"Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009). Imagenet: A large-scale hierarchical image database (pp. 248\u2013255). Ieee."},{"issue":"3","key":"2688_CR8","doi-asserted-by":"publisher","first-page":"1","DOI":"10.1145\/3450494","volume":"2","author":"S Dhar","year":"2021","unstructured":"Dhar, S., Guo, J., Liu, J., Tripathi, S., Kurup, U., & Shah, M. (2021). A survey of on-device machine learning: An algorithms and learning theory perspective. ACM Transactions on Internet of Things,2(3), 1\u201349.","journal-title":"ACM Transactions on Internet of Things"},{"key":"2688_CR9","unstructured":"Dong, C., Loy, C.\u00a0C., He, K., & Tang, X. (2015). Image super-resolution using deep convolutional networks."},{"issue":"1","key":"2688_CR10","first-page":"1997","volume":"20","author":"T Elsken","year":"2019","unstructured":"Elsken, T., Metzen, J. H., & Hutter, F. (2019). Neural architecture search: A survey. The Journal of Machine Learning Research,20(1), 1997\u20132017.","journal-title":"The Journal of Machine Learning Research"},{"issue":"1","key":"2688_CR11","doi-asserted-by":"publisher","first-page":"98","DOI":"10.1007\/s11263-014-0733-5","volume":"111","author":"M Everingham","year":"2015","unstructured":"Everingham, M., Eslami, S. M. A., Van Gool, L., Williams, C. K. I., Winn, J., & Zisserman, A. (2015). The pascal visual object classes challenge: A retrospective. International Journal of Computer Vision,111(1), 98\u2013136.","journal-title":"International Journal of Computer Vision"},{"key":"2688_CR12","doi-asserted-by":"crossref","unstructured":"Gholami, A., Kim, S., Dong, Z., Yao, Z., Mahoney, M.\u00a0W., & Keutzer, K. (2021). A survey of quantization methods for efficient neural network inference.","DOI":"10.1201\/9781003162810-13"},{"key":"2688_CR13","doi-asserted-by":"publisher","first-page":"291","DOI":"10.1201\/9781003162810-13","volume-title":"Low-Power Computer Vision","author":"A Gholami","year":"2022","unstructured":"Gholami, A., Kim, S., Dong, Z., Yao, Z., Mahoney, M. W., & Keutzer, K. (2022). A survey of quantization methods for efficient neural network inference. Low-Power Computer Vision (pp. 291\u2013326). Chapman and Hall\/CRC."},{"key":"2688_CR14","unstructured":"Han, S., Mao, H., & Dally, W.\u00a0J. (2015). Deep compression: Compressing deep neural networks with pruning, trained quantization and huffman coding. arXiv preprint arXiv:1510.00149."},{"key":"2688_CR15","doi-asserted-by":"publisher","DOI":"10.1016\/j.knosys.2020.106622","volume":"212","author":"X He","year":"2021","unstructured":"He, X., Zhao, K., & Chu, X. (2021). Automl: A survey of the state-of-the-art. Knowledge-Based Systems,212, Article 106622.","journal-title":"Knowledge-Based Systems"},{"key":"2688_CR16","doi-asserted-by":"crossref","unstructured":"Hershey, S., Chaudhuri, S., Ellis, D. P. W., Gemmeke, J. F., Jansen, A., Moore, R. C., Plakal, M., Platt, D., Saurous, R. A., Seybold, B., Slaney, M., Weiss, R. J., & Wilson, K. (2017). Cnn architectures for large-scale audio classification. (pp. 131\u2013135)","DOI":"10.1109\/ICASSP.2017.7952132"},{"key":"2688_CR17","doi-asserted-by":"publisher","first-page":"3854","DOI":"10.1609\/aaai.v33i01.33013854","volume":"33","author":"K Huang","year":"2019","unstructured":"Huang, K., Ni, B., & Yang, X. (2019). Efficient quantization for neural networks with binary weights and low bitwidth activations. In Proceedings of the AAAI Conference on Artificial Intelligence,33, 3854\u20133861.","journal-title":"In Proceedings of the AAAI Conference on Artificial Intelligence"},{"key":"2688_CR18","doi-asserted-by":"crossref","unstructured":"Jacob, B., Kligys, S., Chen, B., Zhu, M., Tang, M., Howard, A., Adam, H., & Kalenichenko, D. (2018). Quantization and training of neural networks for efficient integer-arithmetic-only inference.","DOI":"10.1109\/CVPR.2018.00286"},{"key":"2688_CR19","doi-asserted-by":"crossref","unstructured":"Kumari, S., Roy, D., Cartwright, M., Bello, J. P., & Arora, A. (2019). Edgel$${\\hat{\\phantom{a}}}$$3: Compressing l$${\\hat{\\phantom{a}}}$$3-net for mote scale urban noise monitoring. (pp. 877\u2013884)","DOI":"10.1109\/IPDPSW.2019.00145"},{"key":"2688_CR20","doi-asserted-by":"publisher","first-page":"370","DOI":"10.1016\/j.neucom.2021.07.045","volume":"461","author":"T Liang","year":"2021","unstructured":"Liang, T., Glossner, J., Wang, L., Shi, S., & Zhang, X. (2021). Pruning and quantization for deep neural network acceleration: A survey. Neurocomputing,461, 370\u2013403.","journal-title":"Neurocomputing"},{"key":"2688_CR21","doi-asserted-by":"crossref","unstructured":"Liberis, E., Dudziak, \u0141, & Lane, N. D. (2021). $$mu $$nas: Constrained neural architecture search for microcontrollers. (pp. 70\u201379)","DOI":"10.1145\/3437984.3458836"},{"key":"2688_CR22","unstructured":"Lin, J., Chen, W.-M., Cai, H., Gan, C., & Han, S. (2021). Mcunetv2: Memory-efficient patch-based inference for tiny deep learning. arXiv preprint arXiv:2110.15352."},{"key":"2688_CR23","first-page":"11711","volume":"33","author":"J Lin","year":"2020","unstructured":"Lin, J., Chen, W.-M., Lin, Y., Gan, C., Han, S., et al. (2020). Mcunet: Tiny deep learning on iot devices. Advances in Neural Information Processing Systems,33, 11711\u201311722.","journal-title":"Advances in Neural Information Processing Systems"},{"key":"2688_CR24","first-page":"87","volume":"6","author":"J Lin","year":"2024","unstructured":"Lin, J., Tang, J., Tang, H., Yang, S., Chen, W.-M., Wang, W.-C., Xiao, G., Dang, X., Gan, C., & Han, S. (2024). Awq: Activation-aware weight quantization for on-device llm compression and acceleration. Proceedings of Machine Learning and Systems,6, 87\u2013100.","journal-title":"Proceedings of Machine Learning and Systems"},{"key":"2688_CR25","unstructured":"Lin, J., Zhu, L., Chen, W.-M., Wang, W.-C., Gan, C., & Han, S. (2024b). On-device training under 256kb memory."},{"key":"2688_CR26","doi-asserted-by":"crossref","unstructured":"Lin, T.-Y., Maire, M., Belongie, S., Bourdev, L., Girshick, R., Hays, J., Perona, P., Ramanan, D., Zitnick, C.\u00a0L., & Doll\u00e1r, P. (2015). Microsoft coco: Common objects in context.","DOI":"10.1007\/978-3-319-10602-1_48"},{"key":"2688_CR27","doi-asserted-by":"crossref","unstructured":"Liu, W., Anguelov, D., Erhan, D., Szegedy, C., Reed, S., Fu, C.-Y., & Berg, A.\u00a0C. (2016). Ssd: Single shot multibox detector. In Computer Vision\u2013ECCV 2016: 14th European Conference, Amsterdam, The Netherlands, October 11\u201314, 2016, Proceedings, Part I 14, pages 21\u201337. Springer.","DOI":"10.1007\/978-3-319-46448-0_2"},{"key":"2688_CR28","doi-asserted-by":"crossref","unstructured":"Liu, Z., Luo, P., Wang, X., & Tang, X. (2015). Deep learning face attributes in the wild.","DOI":"10.1109\/ICCV.2015.425"},{"key":"2688_CR29","first-page":"1","volume-title":"Ismir","author":"B Logan","year":"2000","unstructured":"Logan, B., et al. (2000). Mel frequency cepstral coefficients for music modeling. Ismir (Vol. 270, pp. 1\u201311). MA: Plymouth."},{"key":"2688_CR30","unstructured":"Magnitskaya, M., Chtchelkatchev, N., Tsvyashchenko, A., Salamatin, D., Lepeshkin, S., Fomicheva, L., & Budzy\u0144ski, M. (2017). Electron and phonon properties of noncentrosymmetric rhge from ab initio calculations."},{"key":"2688_CR31","doi-asserted-by":"publisher","DOI":"10.1016\/j.patcog.2022.109025","volume":"133","author":"M Mohaimenuzzaman","year":"2023","unstructured":"Mohaimenuzzaman, M., Bergmeir, C., West, I., & Meyer, B. (2023). Environmental sound classification on the edge: A pipeline for deep acoustic networks on extremely resource-constrained devices. Pattern Recognition,133, Article 109025.","journal-title":"Pattern Recognition"},{"key":"2688_CR32","unstructured":"Palanisamy, K., Singhania, D., & Yao, A. (2020). Rethinking cnn models for audio classification."},{"key":"2688_CR33","volume-title":"Esc: Dataset for environmental sound classification","author":"KJ Piczak","year":"2015","unstructured":"Piczak, K. J. (2015). Esc: Dataset for environmental sound classification. New York, NY, USA: Association for Computing Machinery."},{"key":"2688_CR34","doi-asserted-by":"crossref","unstructured":"Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection.","DOI":"10.1109\/CVPR.2016.91"},{"key":"2688_CR35","doi-asserted-by":"crossref","unstructured":"Salamon, J., Jacoby, C., & Bello, J. P. (2014). A dataset and taxonomy for urban sound research. 22nd ACM International Conference on Multimedia (ACM-MM\u201914) (pp. 1041\u20131044). FL, USA: Orlando.","DOI":"10.1145\/2647868.2655045"},{"key":"2688_CR36","doi-asserted-by":"crossref","unstructured":"Sandler, M., Howard, A., Zhu, M., Zhmoginov, A., & Chen, L.-C. (2018). Mobilenetv2: Inverted residuals and linear bottlenecks. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 4510\u20134520.","DOI":"10.1109\/CVPR.2018.00474"},{"key":"2688_CR37","unstructured":"STMicroelectronics (2024a). STM32F. https:\/\/www.st.com\/en\/microcontrollers-microprocessors\/stm32f412.html."},{"key":"2688_CR38","unstructured":"STMicroelectronics (2024b). STM32H7. https:\/\/www.st.com\/en\/microcontrollers-microprocessors\/stm32h7-series.html."},{"key":"2688_CR39","doi-asserted-by":"crossref","unstructured":"Tan, M., Chen, B., Pang, R., Vasudevan, V., Sandler, M., Howard, A., & Le, Q. V. (2019). Mnasnet: Platform-aware neural architecture search for mobile.","DOI":"10.1109\/CVPR.2019.00293"},{"key":"2688_CR40","unstructured":"Tan, M., & Le, Q. (2019). In K. Chaudhuri & R. Salakhutdinov (Eds.), EfficientNet: Rethinking model scaling for convolutional neural networks (pp. 6105\u20136114). PMLR."},{"key":"2688_CR41","doi-asserted-by":"publisher","first-page":"63280","DOI":"10.1109\/ACCESS.2022.3182659","volume":"10","author":"S Vadera","year":"2022","unstructured":"Vadera, S., & Ameen, S. (2022). Methods for pruning deep neural networks. IEEE Access,10, 63280\u201363300.","journal-title":"IEEE Access"},{"key":"2688_CR42","doi-asserted-by":"crossref","unstructured":"Wei, L., Ma, Z., Yang, C., & Yao, Q. (2024). Advances in the neural network quantization: A comprehensive review. Applied Sciences,14(17), 7445.","DOI":"10.3390\/app14177445"},{"key":"2688_CR43","unstructured":"Wyse, L. (2017). Audio spectrogram representations for processing with convolutional neural networks. arXiv preprint arXiv:1706.09559."},{"key":"2688_CR44","doi-asserted-by":"crossref","unstructured":"Xu, S., Huang, A., Chen, L., & Zhang, B. (2020). Convolutional neural network pruning: A survey (pp. 7458\u20137463). IEEE.","DOI":"10.23919\/CCC50068.2020.9189610"},{"key":"2688_CR45","unstructured":"Yu, F., Seff, A., Zhang, Y., Song, S., Funkhouser, T., & Xiao, J. (2016). Lsun: Construction of a large-scale image dataset using deep learning with humans in the loop."},{"key":"2688_CR46","doi-asserted-by":"crossref","unstructured":"Zhou, B., Zhao, H., Puig, X., Xiao, T., Fidler, S., Barriuso, A., & Torralba, A. (2018). Semantic understanding of scenes through the ade20k dataset.","DOI":"10.1007\/s11263-018-1140-0"},{"key":"2688_CR47","doi-asserted-by":"crossref","unstructured":"Zoph, B., Vasudevan, V., Shlens, J., & Le, Q.\u00a0V. (2018). Learning transferable architectures for scalable image recognition. In Proceedings of the IEEE conference on computer vision and pattern recognition, pages 8697\u20138710.","DOI":"10.1109\/CVPR.2018.00907"}],"container-title":["International Journal of Computer Vision"],"original-title":[],"language":"en","link":[{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11263-025-02688-w.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/article\/10.1007\/s11263-025-02688-w","content-type":"text\/html","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/link.springer.com\/content\/pdf\/10.1007\/s11263-025-02688-w.pdf","content-type":"application\/pdf","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2026,3,27]],"date-time":"2026-03-27T08:39:59Z","timestamp":1774600799000},"score":1,"resource":{"primary":{"URL":"https:\/\/link.springer.com\/10.1007\/s11263-025-02688-w"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2026,1,29]]},"references-count":47,"journal-issue":{"issue":"3","published-print":{"date-parts":[[2026,3]]}},"alternative-id":["2688"],"URL":"https:\/\/doi.org\/10.1007\/s11263-025-02688-w","relation":{},"ISSN":["0920-5691","1573-1405"],"issn-type":[{"value":"0920-5691","type":"print"},{"value":"1573-1405","type":"electronic"}],"subject":[],"published":{"date-parts":[[2026,1,29]]},"assertion":[{"value":"30 April 2025","order":1,"name":"received","label":"Received","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"17 September 2025","order":2,"name":"accepted","label":"Accepted","group":{"name":"ArticleHistory","label":"Article History"}},{"value":"29 January 2026","order":3,"name":"first_online","label":"First Online","group":{"name":"ArticleHistory","label":"Article History"}}],"article-number":"91"}}