Timezone: »
Tiny deep learning on microcontroller units (MCUs) is challenging due to the limited memory size. We find that the memory bottleneck is due to the imbalanced memory distribution in convolutional neural network (CNN) designs: the first several blocks have an order of magnitude larger memory usage than the rest of the network. To alleviate this issue, we propose a generic patch-by-patch inference scheduling, which operates only on a small spatial region of the feature map and significantly cuts down the peak memory. However, naive implementation brings overlapping patches and computation overhead. We further propose receptive field redistribution to shift the receptive field and FLOPs to the later stage and reduce the computation overhead. Manually redistributing the receptive field is difficult. We automate the process with neural architecture search to jointly optimize the neural architecture and inference scheduling, leading to MCUNetV2. Patch-based inference effectively reduces the peak memory usage of existing networks by4-8×. Co-designed with neural networks, MCUNetV2 sets a record ImageNetaccuracy on MCU (71.8%) and achieves >90% accuracy on the visual wake words dataset under only 32kB SRAM. MCUNetV2 also unblocks object detection on tiny devices, achieving 16.9% higher mAP on Pascal VOC compared to the state-of-the-art result. Our study largely addressed the memory bottleneck in tinyML and paved the way for various vision applications beyond image classification.
Author Information
Ji Lin (MIT)
Wei-Ming Chen (MIT)
Han Cai (Massachusetts Institute of Technology)
Chuang Gan (MIT-IBM Watson AI Lab)
Song Han (MIT)
More from the Same Authors
-
2021 : ThreeDWorld: A Platform for Interactive Multi-Modal Physical Simulation »
Chuang Gan · Jeremy Schwartz · Seth Alter · Damian Mrowca · Martin Schrimpf · James Traer · Julian De Freitas · Jonas Kubilius · Abhishek Bhandwaldar · Nick Haber · Megumi Sano · Kuno Kim · Elias Wang · Michael Lingelbach · Aidan Curtis · Kevin Feigelis · Daniel Bear · Dan Gutfreund · David Cox · Antonio Torralba · James J DiCarlo · Josh Tenenbaum · Josh McDermott · Dan Yamins -
2021 : STAR: A Benchmark for Situated Reasoning in Real-World Videos »
Bo Wu · Shoubin Yu · Zhenfang Chen · Josh Tenenbaum · Chuang Gan -
2021 Poster: Dynamic Visual Reasoning by Learning Differentiable Physics Models from Video and Language »
Mingyu Ding · Zhenfang Chen · Tao Du · Ping Luo · Josh Tenenbaum · Chuang Gan -
2021 Poster: PTR: A Benchmark for Part-based Conceptual, Relational, and Physical Reasoning »
Yining Hong · Li Yi · Josh Tenenbaum · Antonio Torralba · Chuang Gan -
2021 Poster: When does Contrastive Learning Preserve Adversarial Robustness from Pretraining to Finetuning? »
Lijie Fan · Sijia Liu · Pin-Yu Chen · Gaoyuan Zhang · Chuang Gan -
2021 Poster: Delayed Gradient Averaging: Tolerate the Communication Latency for Federated Learning »
Ligeng Zhu · Hongzhou Lin · Yao Lu · Yujun Lin · Song Han -
2021 : ThreeDWorld: A Platform for Interactive Multi-Modal Physical Simulation »
Chuang Gan · Jeremy Schwartz · Seth Alter · Damian Mrowca · Martin Schrimpf · James Traer · Julian De Freitas · Jonas Kubilius · Abhishek Bhandwaldar · Nick Haber · Megumi Sano · Kuno Kim · Elias Wang · Michael Lingelbach · Aidan Curtis · Kevin Feigelis · Daniel Bear · Dan Gutfreund · David Cox · Antonio Torralba · James J DiCarlo · Josh Tenenbaum · Josh McDermott · Dan Yamins -
2020 Poster: MCUNet: Tiny Deep Learning on IoT Devices »
Ji Lin · Wei-Ming Chen · Yujun Lin · john cohn · Chuang Gan · Song Han -
2020 Spotlight: MCUNet: Tiny Deep Learning on IoT Devices »
Ji Lin · Wei-Ming Chen · Yujun Lin · john cohn · Chuang Gan · Song Han -
2020 Poster: Differentiable Augmentation for Data-Efficient GAN Training »
Shengyu Zhao · Zhijian Liu · Ji Lin · Jun-Yan Zhu · Song Han -
2020 Poster: TinyTL: Reduce Memory, Not Parameters for Efficient On-Device Learning »
Han Cai · Chuang Gan · Ligeng Zhu · Song Han -
2020 : Neurosymbolic Visual Reasoning »
Chuang Gan -
2019 : Hardware-aware Neural Architecture Design for Small and Fast Models: from 2D to 3D »
Song Han -
2019 : Posters and Coffee »
Sameer Kumar · Tomasz Kornuta · Oleg Bakhteev · Hui Guan · Xiaomeng Dong · Minsik Cho · Soeren Laue · Theodoros Vasiloudis · Andreea Anghel · Erik Wijmans · Zeyuan Shang · Oleksii Kuchaiev · Ji Lin · Susan Zhang · Ligeng Zhu · Beidi Chen · Vinu Joseph · Jialin Ding · Jonathan Raiman · Ahnjae Shin · Vithursan Thangarasa · Anush Sankaran · Akhil Mathur · Martino Dazzi · Markus Löning · Darryl Ho · Emanuel Zgraggen · Supun Nakandala · Tomasz Kornuta · Rita Kuznetsova -
2019 Poster: Park: An Open Platform for Learning-Augmented Computer Systems »
Hongzi Mao · Parimarjan Negi · Akshay Narayan · Hanrui Wang · Jiacheng Yang · Haonan Wang · Ryan Marcus · Ravichandra Addanki · Mehrdad Khani Shirkoohi · Songtao He · Vikram Nathan · Frank Cangialosi · Shaileshh Venkatakrishnan · Wei-Hung Weng · Song Han · Tim Kraska · Dr.Mohammad Alizadeh -
2019 Poster: Deep Leakage from Gradients »
Ligeng Zhu · Zhijian Liu · Song Han -
2019 Poster: Cross-channel Communication Networks »
Jianwei Yang · Zhile Ren · Chuang Gan · Hongyuan Zhu · Devi Parikh -
2019 Poster: Visual Concept-Metaconcept Learning »
Chi Han · Jiayuan Mao · Chuang Gan · Josh Tenenbaum · Jiajun Wu -
2019 Poster: Imitation Learning from Observations by Minimizing Inverse Dynamics Disagreement »
Chao Yang · Xiaojian Ma · Wenbing Huang · Fuchun Sun · Huaping Liu · Junzhou Huang · Chuang Gan -
2019 Spotlight: Imitation Learning from Observations by Minimizing Inverse Dynamics Disagreement »
Chao Yang · Xiaojian Ma · Wenbing Huang · Fuchun Sun · Huaping Liu · Junzhou Huang · Chuang Gan -
2019 Poster: Point-Voxel CNN for Efficient 3D Deep Learning »
Zhijian Liu · Haotian Tang · Yujun Lin · Song Han -
2019 Spotlight: Point-Voxel CNN for Efficient 3D Deep Learning »
Zhijian Liu · Haotian Tang · Yujun Lin · Song Han -
2018 : Panel disucssion »
Max Welling · Tim Genewein · Edwin Park · Song Han -
2018 : Prof. Song Han »
Song Han -
2018 : Poster spotlight session. »
Abdullah Salama · Wei-Cheng Chang · Aidan Gomez · Raphael Tang · FUXUN YU · Zhendong Zhang · Yuxin Zhang · Ji Lin · Stephen Tiedemann · Kun Bai · Sivaramakrishnan Sankarapandian · Marton Havasi · Jack Turner · Hsin-Pai Cheng · Yue Wang · Xiaofan Xu · Ruizhou Ding · Haoji Hu · Mohammad Shafiee · Christopher Blake · Chieh-Chi Kao · Daniel Kang · Yew Ken Chia · Amir Ashouri · Sourya Basu · Simon Wiedemann · Thorsten Laude -
2018 : Bandwidth efficient deep learning by model compression »
Song Han -
2018 Poster: Weakly Supervised Dense Event Captioning in Videos »
Xin Wang · Wenbing Huang · Chuang Gan · Jingdong Wang · Wenwu Zhu · Junzhou Huang -
2018 Poster: Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding »
Kexin Yi · Jiajun Wu · Chuang Gan · Antonio Torralba · Pushmeet Kohli · Josh Tenenbaum -
2018 Spotlight: Neural-Symbolic VQA: Disentangling Reasoning from Vision and Language Understanding »
Kexin Yi · Jiajun Wu · Chuang Gan · Antonio Torralba · Pushmeet Kohli · Josh Tenenbaum -
2017 Poster: Runtime Neural Pruning »
Ji Lin · Yongming Rao · Jiwen Lu · Jie Zhou