if you want to use a model in detectron2 as front end to detect object in your dataset,you can use this repo to quickly try many model while don't neet to change config.yaml
to fit your dataset,only change some parameter in train.py
, and you can get visual result usevisual_2.py
Only need to change some parameter in train.py
- put config file in
config
- put pretrained weight in
pretrain
visual_v1.py
and evalution use same weight path
- change dataset path in
root_davis
- when use
visual_v1
,recomment to setgpu=1
- if want to get else result ,such as
daivs_test
,changevisual_what
invisual_result.py
visual_v1:
- launch visual_v1 in
train.py
- need
config
andweight
visual_v2:
- independent run,change parameter
- need result.jason
- need register dateset