train_eval_inference.md

Training, Evaluation and Inference

The TB-Net model takes as input images of shape (N, 224, 224, 3) and outputs the softmax probabilities as (N, 2), where N is the batch size.

If using the TF checkpoints, here are some useful tensors:

• input tensor: image:0
• logit tensor: resnet_model/final_dense:0
• output tensor: ArgMax:0
• loss tensor: loss/add:0
• label tensor: classification/label:0

If using the provided dataset interface script, modify the CSV paths accordingly, to point at where the files 'train_split.csv', 'val_split.csv', and 'test_split.csv' are on your system.

Training

We provide a training script that can be used for model training using an untrained model. We provide an untrained version of TB-Net here.

Example command:

python3 train_tbnet.py \
    --weightspath 'TB-Net' \
    --metaname model_train.meta \
    --ckptname model \
    --datapath 'data/' \
    --epochs 10 

Evaluation

We provide an evaluation script that can be used to evaluate a model on the test set. The TB-Net model can be found here.

Example command:

python3 eval.py \
    --weightspath 'TB-Net' \
    --metaname model_eval.meta \
    --ckptname model \
    --datapath 'data/' \

Inference

DISCLAIMER: Do not use this prediction for self-diagnosis. You should check with your local authorities for the latest advice on seeking medical assistance.

We provide an inference script that can be used to evaluate a single image using the given model. The TB-Net model can be found here.

Example command:

python3 inference.py \
    --weightspath 'TB-Net' \
    --metaname model_eval.meta \
    --ckptname model \
    --inputpath 'example_inputs/'