Skip to content

Latest commit

 

History

History
39 lines (25 loc) · 1.6 KB

using_tensorflow.md

File metadata and controls

39 lines (25 loc) · 1.6 KB
Raw

Using TensorFlow

Single Image (TensorFlow)

We have ported our models to TensorFlow to facilitate usage by other teams.

For the cancer classification model, the model has been ported entirely to TensorFlow within this repository.

For the patch classification model, we require the DenseNet code from here. Simply git clone the linked repository (e.g. to /path/to/tensorflow-densenet) and add the path to your PYTHONPATH.

export PYTHONPATH=/path/to/tensorflow-densenet:$PYTHONPATH

Both the cancer classification models and patch classification models are called with the same Python scripts as the PyTorch versions, except with a _tf suffix (e.g. run_models_single_tf.py). See run_single_tf.sh for a full TensorFlow-based pipeline.

In addition, the cancer classification models and patch classification models are also independent and interchangeable with their PyTorch equivalents, so you can use e.g. the PyTorch-generated heatmaps with the TensorFlow classifier.

Additional Prerequisites

Running

Using run_single_tf.sh, you should get the following output:

bash run_single_tf.sh "sample_data/images/0_L_CC.png" "L-CC"
Stage 4a: Run Classifier (Image)
{"benign": 0.04019179195165634, "malignant": 0.008045281283557415}
Stage 4b: Run Classifier (Image+Heatmaps)
{"benign": 0.05236595869064331, "malignant": 0.005510156974196434}

The outputs should be nearly identical to those of the PyTorch models.