Data Free Quantization

Data-Free Quantization is a technique that achieves high-precision quantized models without accessing any training data. This method effectively addresses the issue of data unavailability during the quantization process by generating synthetic boundary support samples, significantly enhancing the deployment efficiency and resource utilization of models, particularly in the field of computer vision optimization and acceleration.