Object Detection Steps and Resources

Limelight documentation for training a model

• Read through this for steps
• Look @ Roboflow Universe for models
• Figure out how to combine them??

Where to upload the model - once you have all your annotated date

1. What is object detection & how have we used it in the past
2. Find models on RoboFlow Universe
   1. What makes a good model
      1. Variety of images - different contexts
3. Converting COCO JSON to Tensorflow TF Record
   1. Step 1: set up the project
   2. Step 2: upload your data to the project
      1. Download the models you found as .coco files
      2. Upload data to the model → move in the coco files & upload them to the dataset
   3. Step 3 dataset version
      1. Auto orient
      2. Resize
      3. grayscale
      4. 
   4. Step 4
      1. Add 90 degree rotation
      2. Add + & - 15 degree brightness
   5. Download tensorflow tf record
4. Now you’ve created a new version
5. Upload final TF Record model <- TO THIS LINK
   1. Train the model (takes some time)
   2. When done, you get a tf.white file
6. Add to the code - follow Upload to Limelight step HERE
7. 7.5 Detector Results → use API call to grab the location information
8. Look HERE (2024 Crescendo) for an example of a Game Piece Alignment Command