Answer Posted / Astha Tulsyan

To retrain an image classifier for new categories using transfer learning, follow these steps:

1. Find a pre-trained model (e.g., Inception V3) on ImageNet dataset with the desired number of output classes.

2. Download the weights of the pre-trained model and extract the feature extractor (without the final fully connected layers).

3. Prepare new labeled data for the target categories, resizing them to match the input size of the feature extractor.

4. Use the feature extractor to get feature vectors from the new data.

5. Fine-tune a small number of fully connected layers on top of the feature extractor using your labeled data.

6. Train the model with the fine-tuned layers and evaluate its performance on validation data.

What are the apis outside tensorflow project?

4