Use Case and High-Level Description

This is the Resnet-50 v1 model that is designed to perform image classification. The model has been pretrained on the ImageNet image database and then quantized to INT8 fixed-point precision using so-called Quantization-aware training approach implemented in TensorFlow framework. For details about the original floating point model, check out the paper.

The model input is a blob that consists of a single image of "1x224x224x3" in BGR order.

The model output for resnet-50-int8-tf-0001 is the usual object classifier output for the 1000 different classifications matching those in the ImageNet database.



Metric Value
Type Classification
GFLOPs 6.996
MParams 25.530
Source framework TensorFlow


The quality metrics calculated on ImageNet validation dataset is 74.6% accuracy top-1.

Metric Value
Accuracy top-1 (ImageNet) 74.6%



Image, shape - 1,224,224,3, format is B,H,W,C where:

Channel order is BGR


Object classifier according to ImageNet classes, shape -1,1000, output data format is B,C where: