# Designing Extremely Memory-Efficient CNNs for On-device Vision Tasks

arXiv: https://arxiv.org/abs/2408.03663 · HTML: https://mlsanity.com/paper/2408.03663/

## Key facts

- Cited by 2 papers in the map (data through 14 Oct 2026).
- Builds on 19 of its 30 references that are in the map.
- First posted on arXiv 7 Aug 2024.

## Builds on (most cited first, 19 of 19)

- [Microsoft COCO: Common Objects in Context](https://mlsanity.com/paper/1405.0312/) (2014)
- [MobileNetV2: Inverted Residuals and Linear Bottlenecks](https://mlsanity.com/paper/1801.04381/) (2018)
- [The Cityscapes Dataset for Semantic Urban Scene Understanding](https://mlsanity.com/paper/1604.01685/) (2016)
- [Deep Learning Face Attributes in the Wild](https://mlsanity.com/paper/1411.7766/) (2014)
- [SSD: Single Shot MultiBox Detector](https://mlsanity.com/paper/1512.02325/) (2015)
- [Deep Compression: Compressing Deep Neural Networks with Pruning, Trained Quantization and Huffman Coding](https://mlsanity.com/paper/1510.00149/) (2015)
- [LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loop](https://mlsanity.com/paper/1506.03365/) (2015)
- [AWQ: Activation-aware Weight Quantization for LLM Compression and Acceleration](https://mlsanity.com/paper/2306.00978/) (2023)
- [Image Super-Resolution Using Deep Convolutional Networks](https://mlsanity.com/paper/1501.00092/) (2014)
- [Learning Transferable Architectures for Scalable Image Recognition](https://mlsanity.com/paper/1707.07012/) (2017)
- [Semantic Understanding of Scenes through the ADE20K Dataset](https://mlsanity.com/paper/1608.05442/) (2016)
- [A Survey of Quantization Methods for Efficient Neural Network Inference](https://mlsanity.com/paper/2103.13630/) (2021)
- [AutoML: A Survey of the State-of-the-Art](https://mlsanity.com/paper/1908.00709/) (2019)
- [MCUNet: Tiny Deep Learning on IoT Devices](https://mlsanity.com/paper/2007.10319/) (2020)
- [Pruning and Quantization for Deep Neural Network Acceleration: A Survey](https://mlsanity.com/paper/2101.09671/) (2021)
- [MCUNetV2: Memory-Efficient Patch-based Inference for Tiny Deep Learning](https://mlsanity.com/paper/2110.15352/) (2021)
- [On-Device Machine Learning: An Algorithms and Learning Theory Perspective](https://mlsanity.com/paper/1911.00623/) (2019)
- [μNAS: Constrained Neural Architecture Search for Microcontrollers](https://mlsanity.com/paper/2010.14246/) (2020)
- [Methods for Pruning Deep Neural Networks](https://mlsanity.com/paper/2011.00241/) (2020)

## Built on by (most cited first, 2 of 2)

- [A Survey of Small Language Models](https://mlsanity.com/paper/2410.20011/) (2024)
- [From Tiny Machine Learning to Tiny Deep Learning: A Survey](https://mlsanity.com/paper/2506.18927/) (2025)

## Abstract (arXiv)

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 tasks, such as image classification and object detection, using extremely low memory, i.e., 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, 'input segmentation' 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, 'patch tunneling' 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, 'bottleneck reordering' 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. The experiment result shows that the proposed network classifies ImageNet with extremely low memory (i.e., 63 KB) while achieving competitive top-1 accuracy (i.e., 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, i.e., up to 89x and 3.1x smaller than MobileNet (i.e., 5.6 MB) and MCUNet (i.e., 196 KB), respectively.