White Papers
11 CheXNet – Inference with Nvidia T4 on Dell EMC PowerEdge R7425
3 Development Methodology
In this section we explain the general instructions on how we trained the custom model
CheXNet from scratch with TensorFlow framework using transfer Learning, and how the
trained model was optimized then with TensorRTâ„¢ to run accelerated inferencing.
3.1 Build a CheXNet Model with TensorFlow Framework
The CheXNet model was developed using transfer Learning based on resnet_v2_50, it
means we built the model using the TensorFlow official pre-trained resnetV2_50 checkpoints
downloaded from its website. The model was trained with 14 output classes representing the
thoracic deceases.
In the next paragraphs and snippet codes we will explain the steps and the APIs used to
build the model. Figure 5 shows the general workflow pipeline followed:
Figure 5: Training workload of the custom model CheXNet
Define the Classes:
Below is listed the 14 distinct categories of thoracic diseases to be predicted for the multiclass
classification model
classes = ['Cardiomegaly',
'Emphysema',