pretrained_model and checkpoints parameters in det_mv3_db.yml in detection training

[anidiatm41]

Experts , I am custom training the DB detection model using my own annotated data .

While setting the yml config file (det_mv3_db.yml), I came across below two parameters :

• pretrained_model:
• checkpoints:

I want to know,

Q1. what is the use of this two parameters? Q2. If I want to do a training from scratch..what would be the parameters value? Q3. If I want to do a transfer learning from last checkpoints ...what will be the parameters value?

[anidiatm41]

@andyjpaddle

[anidiatm41]

@Evezerest

[MohieEldinMuhammad]

@anidiatm41 if you knew how to continue training on the trained model let me know if you please

[MohieEldinMuhammad]

i have noticed that train.py is starting the arch from scratch i need to do transfer learning