Adding additional class to tiny yolo

thtrieu / darkflow

Translate darknet to tensorflow. Load trained weights, retrain/fine-tune using tensorflow, export constant graph def to mobile devices

GNU General Public License v3.0

6.13k stars 2.08k forks source link

Adding additional class to tiny yolo #537

Open kumarvvin opened 6 years ago

kumarvvin commented 6 years ago

Hi,

The current tiny-yolo.cfg identifies certain classes, how to add my custom class to it. I want to identify person and my custom trained object in the same model. when i trained new model with own data it works but i want both person and custom object to be detected.

Regards Vaibav

SkullPirateTK commented 6 years ago

Instructions are given in readme.md.

alokrajiv commented 6 years ago

I am not sure its very apparent in the readme here or at darknet. I think this question is regarding preserving existing classes and continue training incrementally for transfer learning. There is very little articles or docs on this procedure atm.

The cfg update option mentioned works only if you start training from scratch with voc dataset + new custom dataset.

Atleast this is what I understand. Please correct me if I'm wrong. (I would be happier to be wrong :D )