Closed terryII closed 3 months ago
hi,感谢你们开创性的工作,我注意到在cogcom中的steps中包含grounding,crop_and_zoomin,counting,OCR等func操作,根据项目需要,我希望在里面添加可以检测关键点的fun操作,如pose操作,其形式为[x1,y1,x2,y2,kpt1,kpt2],其中,kpt1和kpt2为目标的关键点坐标,请问添加关键点func操作除在生成com数据时需要进行一定的修改,在finetuning时是否需要修改finetuning部分的代码? @erjanmx @Sleepychord @cenyk1230 @Btlmd @1049451037
grounding,crop_and_zoomin,counting,OCR
pose
[x1,y1,x2,y2,kpt1,kpt2]
kpt1
kpt2
Hit @terryII , 感谢关注。添加你描述的新操作只需生成对应的数据,无需修改其他代码。
hi,感谢你们开创性的工作,我注意到在cogcom中的steps中包含
grounding,crop_and_zoomin,counting,OCR
等func操作,根据项目需要,我希望在里面添加可以检测关键点的fun操作,如pose
操作,其形式为[x1,y1,x2,y2,kpt1,kpt2]
,其中,kpt1
和kpt2
为目标的关键点坐标,请问添加关键点func操作除在生成com数据时需要进行一定的修改,在finetuning时是否需要修改finetuning部分的代码? @erjanmx @Sleepychord @cenyk1230 @Btlmd @1049451037