search

ermig1979 / Synet

A small framework to infer neural network
MIT License
140 stars 26 forks source link

Instruction for model training #22

Open edward9112 opened 3 years ago

edward9112 commented 3 years ago

Hello Igor!

What do you think about making a brief instruction on how to annotate and train models to be compatible with Synet? Which software to use for annotation, which for training, general recommendations/requirements, etc.

It would be awesome to have that.

Thank you!

ermig1979 commented 3 years ago

Hello!

It's a good idea. I will add it to plans.

edward9112 commented 3 years ago

Hello!

It's a good idea. I will add it to plans.

Awesome!

edward9112 commented 2 years ago

Hey! Any updates on this?

ermig1979 commented 2 years ago

I am sorry but there is no update still. I have got no free time to do this issue.

edward9112 commented 2 years ago

I am sorry but there is no update still. I have got no free time to do this issue.

Thanks for the update. Waiting patiently…

edward9112 commented 2 years ago

Hello! Any updates on the tutorial? Thank you!

edward9112 commented 2 years ago

Just a friendly follow-up @ermig1979

ermig1979 commented 2 years ago

Hi! I see this issue is in demand. Why it does not move? One hand I don't like to make tutorial like other hand there is no definite rules to make models compatible with Synet. I try step by step to extend amount of compatible model types. If I see incompatibility I add the model (in OpenVINO or ONNX format) to test framework and achieve then of correct conversion to internal Synet format. The process still go on.

edward9112 commented 2 years ago

Right, but you probably do some basic model training internally in your team according to some standard procedure, i.e. Step 1: use such and such software for data labeling Step 2: run such and such command on labeled data to retrieve any kind of viable model etc.

This information would be really helpful for those who are not very proficient with model training, and it would be a great addition to your awesome Synet framework.