FORGE: Forward-Only Test-Time Adaptation for Integer-Only Vision Models on Microcontrollers

Researchers introduced FORGE, a forward-only test-time adaptation method for integer-only vision models on microcontrollers, recovering 93% of gradient-based TENT's accuracy gain (+20.9 vs. +24.9 points) while being the only method that runs on folded integer-only models. Deployed on an ESP32-S3, adaptation costs only 8.3 mJ (6.8% of inference energy) and 21.9 ms, demonstrating feasibility on real hardware.

arXiv:2609.01683v1 Announce Type: new Abstract: Vision models deployed on microcontrollers MCUs are quantized to integer-only arithmetic and run in inference-only runtimes that do not carry the machinery backpropagation needs: the standard tool for adapting a model to the distribution shift sensor noise, blur, lighting it meets in the field. Existing forward-only test-time adaptation TTA methods either run only on server- or edge-GPU-class models not true microcontroller integer execution , or require the batch-normalization BN layers that integer deployment fuses away. We present a forward-only TTA method that operates on deployed, BN-folded, integer-only convolutional networks. The key observation is that fusing BN into the preceding convolution, a mandatory step for integer inference, destroys the statistics that normalization-based adaptation relies on. We restore adaptation by re-normalizing each folded convolution's per-channel output to its clean training statistics, using only forward-pass estimates. The method i recovers most of gradient-based TENT's accuracy gain +20.9 vs. +24.9 points and matches forward-only BN adaptation, while being the only method that runs on a folded integer-only model; ii needs to adapt only 3 of 21 layers selected without seeing the test corruptions to recover 93% of the benefit; iii survives single-sample streaming with a batch-size-scaled momentum; and iv generalizes across three datasets up to 200 classes and two architectures. We validate bit-exact int8 convolution execution and deploy on an ESP32-S3, where, measured with a Nordic PPK2 power profiler, the forward-only adaptation a lightweight fp32 recalibration around the int8 convolutions costs only 8.3 mJ 6.8% of inference energy and 21.9 ms on the deployed SIMD-optimized model: forward-only adaptation is cheap on a real microcontroller.