Reply by: MLPractitioner
200 images is on the low side but you can make it work with transfer learning. Start with a pre-trained model like resnet or mobilenet and fine-tune it on your data. Data augmentation absolutely helps - use rotation, flipping, brightness adjustments etc. Also look into using synthetic data if possible. For electronic components you might be able to render 3d models to generate more training data. That technique works really well for industrial applications.

Username: DataScientist_
Posted: 4 days ago
Need to train a classification model but only have about 200 labeled images. Is this enough or do i need more data? I've heard about data augmentation but not sure if it actually helps or just artificially inflates the dataset. Project is for classifying different types of electronic components.