This repository presents a simple to understand and easy to follow guide of implementing ResNet-50 in TensorFlow. It is recommended to first study ResNet paper avialable here.
Input to the network is fed through TFRecord file, made separately for training and testing.
All the parameters are placed in param.py
1. Clone or donwload the repository on local machine
2. Copy train.tfrecords and test.tfrecords file in the directory
3. Modify model/training parameters in param.py file
4. Run multi_gpu_train.py file to start training
5. After training, run ResNet_eval.py for test/eval