# Data-Free Quantization Through Weight Equalization and Bias Correction

## TL;DR

This paper presents a data-free quantization method for deep neural networks that maintains near-original performance without fine-tuning, using weight equalization and bias correction, applicable to various architectures and tasks.

## Contribution

The authors introduce a novel data-free quantization technique leveraging weight equalization and bias correction, eliminating the need for fine-tuning or hyperparameter tuning.

## Key findings

- Achieves near-original model performance on common architectures.
- Extends effectively to tasks like semantic segmentation and object detection.
- State-of-the-art results on MobileNet family models.

## Abstract

We introduce a data-free quantization method for deep neural networks that does not require fine-tuning or hyperparameter selection. It achieves near-original model performance on common computer vision architectures and tasks. 8-bit fixed-point quantization is essential for efficient inference on modern deep learning hardware. However, quantizing models to run in 8-bit is a non-trivial task, frequently leading to either significant performance reduction or engineering time spent on training a network to be amenable to quantization. Our approach relies on equalizing the weight ranges in the network by making use of a scale-equivariance property of activation functions. In addition the method corrects biases in the error that are introduced during quantization. This improves quantization accuracy performance, and can be applied to many common computer vision architectures with a straight forward API call. For common architectures, such as the MobileNet family, we achieve state-of-the-art quantized model performance. We further show that the method also extends to other computer vision architectures and tasks such as semantic segmentation and object detection.