How to improve Rekognition Content Moderation model?

Hi, Oh, okay. Could you then help me with these questions:

1. If we are not directly uploading videos for training, could I instead use images dataset rather than frames extracted? Could you please explain what is the difference?
2. I see the custom model console uses image classification or object detection on custom labels. Can you tell me more about what are the things to consider while training for custom content moderation dataset of videos? If the training models are for images itself, how will that apply on my videos?

Hey, So custom labels helps you create an image classification/multilabel/objectdetection model inside Amazon Rekognition. You will use an image dataset to train the model. Once the model has been trained, the service will manage the model endpoint for you and you will be able to call Amazon Rekognition to detect labels/objects in your images, using the model you have trained. As Custom Labels does not support videos as input (other rekognition APIs do), you have to split your videos into frames and call the service passing the frame as image, the code example provided above helps you with this workaround! Hope this helps and remember to accept the answer if it has helped you (^_^)