Building models is hard and it can take a lot of trial and error to get them working, and prove that they are robust. We’ve put together this list of tips, based on our experience, to help you build more accurate, less biased models. We typically work with image detection and object classification models, but you may be able to extrapolate some of these points to other domains:

1. Understand Bias

2. Understand your Training Data

3. Get better at Annotating

4. Explore your Results

5. Have your Model Explain itself

6. Recognise Faults early and cut Losses

7. Make your Training Data and results available to your Team

Download our whitepaper to explore each of these steps in further detail.