# FixyNN: Efficient Hardware for Mobile Computer Vision via Transfer Learning

arXiv: https://arxiv.org/abs/1902.11128 · HTML: https://mlsanity.com/paper/1902.11128/

## Key facts

- Cited by 10 papers in the map (data through 14 Oct 2026).
- First posted on arXiv 27 Feb 2019.

## Built on by (most cited first, 10 of 10)

- [hls4ml: An Open-Source Codesign Workflow to Empower Scientific Low-Power Machine Learning Devices](https://mlsanity.com/paper/2103.05579/) (2021)
- [S2TA: Exploiting Structured Sparsity for Energy-Efficient Mobile CNN Acceleration](https://mlsanity.com/paper/2107.07983/) (2021)
- [Open-source FPGA-ML codesign for the MLPerf Tiny Benchmark](https://mlsanity.com/paper/2206.11791/) (2022)
- [Model-to-Circuit Cross-Approximation For Printed Machine Learning Classifiers](https://mlsanity.com/paper/2303.08255/) (2023)
- [Doping: A technique for efficient compression of LSTM models using sparse structured additive matrices](https://mlsanity.com/paper/2102.07071/) (2021)
- [Simple Transferability Estimation for Regression Tasks](https://mlsanity.com/paper/2312.00656/) (2023)
- [Generalization Bounds for Deep Transfer Learning Using Majority Predictor Accuracy](https://mlsanity.com/paper/2209.05709/) (2022)
- [MIME: Adapting a Single Neural Network for Multi-task Inference with Memory-efficient Dynamic Pruning](https://mlsanity.com/paper/2204.05274/) (2022)
- [Drastically Reducing the Number of Trainable Parameters in Deep CNNs by Inter-layer Kernel-sharing](https://mlsanity.com/paper/2210.14151/) (2022)
- [NPU Offloading of a Frozen Visual Encoder for Robot Policy Training](https://mlsanity.com/paper/2608.15002/) (2026)

## Abstract (arXiv)

The computational demands of computer vision tasks based on state-of-the-art Convolutional Neural Network (CNN) image classification far exceed the energy budgets of mobile devices. This paper proposes FixyNN, which consists of a fixed-weight feature extractor that generates ubiquitous CNN features, and a conventional programmable CNN accelerator which processes a dataset-specific CNN. Image classification models for FixyNN are trained end-to-end via transfer learning, with the common feature extractor representing the transfered part, and the programmable part being learnt on the target dataset. Experimental results demonstrate FixyNN hardware can achieve very high energy efficiencies up to 26.6 TOPS/W (4.81 × better than iso-area programmable accelerator). Over a suite of six datasets we trained models via transfer learning with an accuracy loss of <1% resulting in up to 11.2 TOPS/W - nearly 2 × more efficient than a conventional programmable CNN accelerator of the same area.