Final project, the review.
The final project in six concepts and five new questions, from the olive tree case to your own. If an answer doesn't add up, under the explanation you'll find the chapter to read again.
The key concepts.
- Edge AI
Running a model on the device where the data is created, like a phone, a watch or a camera, instead of in a data center.
- Class
One of the possible answers of a model that classifies. Healthy and sick are two classes, and so are walking, running and standing still.
- Dataset
The set of examples you use to train and check a model. Photos, audio clips, sensor recordings or sentences, each one with its own label.
- Models in a row
Several small models working one after the other, each with a narrow task, for example checking the photo, working out what it shows and then giving the diagnosis.
- Quantization
Writing the weights with simpler numbers, for example 8-bit integers instead of 32-bit decimals. The model gets about four times smaller and faster, and it usually loses little.
- Model card
A short page that comes with a model. It says what it's for, what data it was trained on, how it was measured, where it goes wrong and where it shouldn't be used.
5 questions that weren't in the lessons.
They bring the course's lessons together. After each answer there's the explanation and the place to read again. The review doesn't change the map.
-
You want to recognize three plants in the garden, but you can't tell two of them apart yourself. What do you do before you collect the photos?
If you can't label an example, the model can't learn it. Classes are written as a guide for a person, and a model can't be more precise than its classes. Explained in 07.01, chapter 2 →
-
You took the photos for your project on two different days. Which ones do you keep for the test?
The test examples come from a separate session. If two nearly identical shots in a row end up one in training and one in the test, the model looks better than it is. Explained in 01.05, Go deeper →
-
Between one training run and the next, you add new photos and also change the number of epochs. The model gets better. What do you know now?
If you change several things at once, you don't know which one made the model better. Change one at a time between training runs, and write down what you changed. Explained in 07.02, chapter 4 →
-
Your model gives a diagnosis even for blurry or dark photos. What do you add?
In a row of models, the first one checks the photo and throws out blurry or dark shots right away. That way no computation is wasted, and no diagnosis comes from a photo you can't use. Explained in 07.01, Go deeper →
-
Your model is good at spotting mold on the plants in your vegetable garden. What makes it really useful to the person who grows them?
A model answers a narrow question. Its value comes when the answer turns into a concrete decision. OliveSense combines the diagnosis with the weather, the satellite and the treatments, and suggests a dose and a time. Explained in 07.01, chapter 5 →