# AHC: Meta-Learned Adaptive Compression for Continual Object Detection on Memory-Constrained Microcontrollers

arXiv: https://arxiv.org/abs/2604.09576 · HTML: https://mlsanity.com/paper/2604.09576/

## Key facts

- No citing papers in the map yet (data through 14 Oct 2026).
- Builds on 22 of its 28 references that are in the map.
- First posted on arXiv 24 Feb 2026.

## Builds on (most cited first, 22 of 22)

- [EfficientNet: Rethinking Model Scaling for Convolutional Neural Networks](https://mlsanity.com/paper/1905.11946/) (2019)
- [Model-Agnostic Meta-Learning for Fast Adaptation of Deep Networks](https://mlsanity.com/paper/1703.03400/) (2017)
- [Overcoming catastrophic forgetting in neural networks](https://mlsanity.com/paper/1612.00796/) (2016)
- [MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications](https://mlsanity.com/paper/1704.04861/) (2017)
- [Learning without Forgetting](https://mlsanity.com/paper/1606.09282/) (2016)
- [iCaRL: Incremental Classifier and Representation Learning](https://mlsanity.com/paper/1611.07725/) (2016)
- [FiLM: Visual Reasoning with a General Conditioning Layer](https://mlsanity.com/paper/1709.07871/) (2017)
- [Gradient Episodic Memory for Continual Learning](https://mlsanity.com/paper/1706.08840/) (2017)
- [FCOS: Fully Convolutional One-Stage Object Detection](https://mlsanity.com/paper/1904.01355/) (2019)
- [Generalized Intersection over Union: A Metric and A Loss for Bounding Box Regression](https://mlsanity.com/paper/1902.09630/) (2019)
- [On First-Order Meta-Learning Algorithms](https://mlsanity.com/paper/1803.02999/) (2018)
- [Dark Experience for General Continual Learning: a Strong, Simple Baseline](https://mlsanity.com/paper/2004.07211/) (2020)
- [Meta-SGD: Learning to Learn Quickly for Few-Shot Learning](https://mlsanity.com/paper/1707.09835/) (2017)
- [Towards Open World Object Detection](https://mlsanity.com/paper/2103.02603/) (2021)
- [CORe50: a New Dataset and Benchmark for Continuous Object Recognition](https://mlsanity.com/paper/1705.03550/) (2017)
- [MCUNet: Tiny Deep Learning on IoT Devices](https://mlsanity.com/paper/2007.10319/) (2020)
- [Rainbow Memory: Continual Learning with a Memory of Diverse Samples](https://mlsanity.com/paper/2103.17230/) (2021)
- [Few-shot Object Detection via Feature Reweighting](https://mlsanity.com/paper/1812.01866/) (2018)
- [Incremental Learning of Object Detectors without Catastrophic Forgetting](https://mlsanity.com/paper/1708.06977/) (2017)
- [REMIND Your Neural Network to Prevent Catastrophic Forgetting](https://mlsanity.com/paper/1910.02509/) (2019)
- [Online Meta-Learning for Multi-Source and Semi-Supervised Domain Adaptation](https://mlsanity.com/paper/2004.04398/) (2020)
- [Continual Detection Transformer for Incremental Object Detection](https://mlsanity.com/paper/2304.03110/) (2023)

## Abstract (arXiv)

Deploying continual object detection on microcontrollers (MCUs) with under 100KB memory requires efficient feature compression that can adapt to evolving task distributions. Existing approaches rely on fixed compression strategies (e.g., FiLM conditioning) that cannot adapt to heterogeneous task characteristics, leading to suboptimal memory utilization and catastrophic forgetting. We introduce Adaptive Hierarchical Compression (AHC), a meta-learning framework featuring three key innovations: (1) true MAML-based compression that adapts via gradient descent to each new task in just 5 inner-loop steps, (2) hierarchical multi-scale compression with scale-aware ratios (8:1 for P3, 6.4:1 for P4, 4:1 for P5) matching FPN redundancy patterns, and (3) a dual-memory architecture combining short-term and long-term banks with importance-based consolidation under a hard 100KB budget. We provide formal theoretical guarantees bounding catastrophic forgetting as O(εsq.root(T) + 1/sq.root(M)) where ε is compression error, T is task count, and M is memory size. Experiments on CORe50, TiROD, and PASCAL VOC benchmarks with three standard baselines (Fine-tuning,EWC, iCaRL) demonstrate that AHC enables practical continual detection within a 100KB replay budget, achieving competitive accuracy through mean-pooled compressed feature replay combined with EWC regularization and feature distillation.