My apologies, I am not fully sure on all the questions. But let me still make an attempt to respond to see if it helps.
Yes, you can write your own custom code through SageMaker studio.
This may not be an apple to apple comparison. The main advantage in this context, is your able to scale out your training to multiple nodes and cores (if your underlying model supports that). Likewise you can scale out the deployment as well. Typically the studio notebook is backed by a lightweight EC2 instance, but there are a large range of EC2 instances for training on SageMaker. Please refer to the following links for further assistance. 1. https://docs.aws.amazon.com/sagemaker/latest/dg/notebooks-available-instance-types.html 2. https://aws.amazon.com/ec2/instance-types/
Please refer to the response above for question # 2.
Did you mean semantic segmentation? If yes, the answer is yes too.
Hope that helps!
Can I use SageMaker Autopilot for my time-series modeling?Accepted Answerasked 2 years ago
how can I use sagemaker_sklearn_extension in Sagemaker job?asked 5 months ago
How to create custom templates for model training and building in sagemaker studio?asked 3 months ago
Exporting Sagemaker model to local computerasked 8 months ago
Host multiple TensorFlow computer vision models using Amazon SageMaker multi-model endpointsasked 10 months ago
How can make multi model endpoint with SageMaker?Accepted Answerasked 4 months ago
Yolo5 model deployment into SageMaker endpointasked 3 months ago
Which Amazon SageMaker algorithms can only use GPU for training?Accepted AnswerMODERATORasked 2 years ago
Ask AWS SageMakerAccepted Answerasked 4 months ago
How to create (Serverless) SageMaker Endpoint using exiting tensorflow pb (frozen model) file?asked 7 months ago