Deep Learning for Computer Vision

Re-training object detection models

The same API lets us retrain a model for our custom dataset. Training of custom data involves the preparation of a dataset, selecting the algorithm, and performing fine-tuning. The whole pipeline can be passed as a parameter to the training script. The training data has to be converted to TensorFlow records. TensorFlow records is a file format provided by Google to make the reading of data faster than regular files. Now, we will go through the steps of training.