densenet-169-tf

Use Case and High-Level Description

This is a TensorFlow* version of densenet-169 model, one of the DenseNet group of models designed to perform image classification. The weights were converted from DenseNet-Keras Models. For details, see repository and paper.

Specification

Metric Value
Type Classification
GFlops 6.16
MParams 14.139
Source framework TensorFlow*

Accuracy

Metric Value
Top 1 75.76%
Top 5 92.81%

Input

Original Model

Name: Placeholder , shape: [1x224x224x3]. An input image is the [BxHxWxC] format, where:

  • B - batch size
  • H - image height
  • W - image width
  • C - number of channels

Expected color order: RGB. Mean values: [123.68, 116.78, 103.94], scale factor for each channel: 58.8235294

Converted Model

Name: Placeholder, shape: [1x3x224x224]. An input image is in the [BxCxHxW] format, where:

  • B - batch size
  • C - number of channels
  • H - image height
  • W - image width

Expected color order: BGR.

Output

Original Model

Name:densenet169/predictions/Reshape_1. Contains floating point values in a range [0, 1], which represent probabilities for classes in a dataset.

Converted Model

Name: densenet169/predictions/Reshape_1/Transpose, shape: [1, 1, 1, 1000]. Contains floating point values in a range [0, 1], which represent probabilities for classes in a dataset.

Legal Information

The original model is distributed under the Apache License, Version 2.0. A copy of the license is provided in APACHE-2.0-TF-DenseNet.txt.