added images

notebooks/community/fantasy_maps/README.md

@@ -0,0 +1,45 @@
+# Fantasy Maps
+
+Virtual tabletop gaming continues to grow in popularity. Often times, a core requirement for
+virtual tabletop gaming apps (VTTs) is to provide a map (an image) and some metadata that 
+describes the valid game moves (as a grid) on that map.
+
+With this explosion in popularity of VTTs, many folks in the gaming community publish their
+own maps for others to use. However, the maps don't always include the metadata needed for
+a VTT app.
+
+That's where the Fantasy Maps project comes in.
+
+This project uses
+[Vertex AI](https://cloud.google.com/vertex-ai),
+AutoML object detection models to train a model capable of identifying gridlines on a map
+like the example below:
+
+![Predictions shown on map of a desert](resources/predicted-gridded-desert.png)
+
+## Products demonstrated
+
+This project demonstrates how to combine multiple Google Cloud products together
+to build this ML model. Specifically, the project demonstrates:
+
++ How to use [Firestore](https://cloud.google.com/firestore/docs) to store Vertex AI
+  training metadata for unstructured data (e.g. images)
++ How to use [Secret Manager](https://cloud.google.com/secret-manager/docs) to keep
+  API keys hidden until they're needed
++ How to store data in [Cloud Storage](https://cloud.google.com/storage/docs)
++ How to build a
+  [Vertex AI Pipeline](https://cloud.google.com/vertex-ai/docs/pipelines/introduction?hl=en)
+  to train, deploy, and retrain a Vertex AI model
++ How to use Vertex AI to to train, deploy, and get inferences from an image object
+  detection model
+
+## Tutorial version
+
+The project consists of five notebooks, where each notebook builds upon the lessons
+of the previous notebook. The notebooks are as follows:
+
+1. [Storing training data in Firestore](1_firestore.ipynb)
+2. Creating a Vertex AI Pipeline to store training data in Firestore
+3. Building an AutoML image object detection model
+4. Creating a Vertex AI Pipeline to train and deploy an AutoML model
+5. Getting predictions from the deployed model