If you can write your own layer and add it to the layers and then fine tune it would be good. Especially using weighted classes to pay more attention to those underrepresented classes. This means you change the optimization algorithm. Then you need to add your own .config file.

Otherwise, a simple solution is to use metrics that are more suitable for imbalanced datasets, like AUC, F1-score, etc.

Dear Isaac Barnhart

These links might be helpful, have a look:

1. https://www.tensorflow.org/tutorials/structured_data/imbalanced_data

2. https://www.tensorflow.org/tutorials/images/transfer_learning

3. https://neptune.ai/blog/tensorflow-object-detection-api-best-practices-to-training-evaluation-deployment

4. https://curiousily.com/posts/practical-guide-to-handling-imbalanced-datasets/

5. https://towardsdatascience.com/4-ways-to-improve-class-imbalance-for-image-data-9adec8f390f1

6. https://stackoverflow.com/questions/51904077/imbalanced-dataset-object-detection

7. https://blog.roboflow.com/training-a-tensorflow-faster-r-cnn-object-detection-model-on-your-own-dataset/

8. https://machinelearningmastery.com/how-to-train-an-object-detection-model-with-keras/

9. https://www.analyticsvidhya.com/blog/2020/07/transfer-learning-for-nlp-fine-tuning-bert-for-text-classification/

Kind Regards

Qamar Ul Islam

Dear Isaac, the following link should be useful:

https://stackoverflow.com/questions/51862997/class-weights-for-balancing-data-in-tensorflow-object-detection-api