tf-ssd
tf-ssd copied to clipboard
TRAINING ON A CUSTOM DATASET
i am wondering what changes are needed to train the model on a custom dataset. thank you
I am wondering what changes are needed to train the model on a custom dataset. thank you
Me too. could you please let me know what are the changes if know? Thanks in advance
You need to change the method preprocessing
in utils/data_utils.py
.
What I did was store my custom dataset in a TFRecord format. Look into how to store and read TFRecords in Tensorflow (it's simple after you understand it).
After that, map the preprocessing method in your dataset (make the same pipeline as the method in this repository), then the rest of the code is basically the same.