Retraining with lesser COCO images.

ZheC / Realtime_Multi-Person_Pose_Estimation

Code repo for realtime multi-person pose estimation in CVPR'17 (Oral)

Other

5.1k stars 1.37k forks source link

Retraining with lesser COCO images. #229

Open Aakriti05 opened 5 years ago

Aakriti05 commented 5 years ago

I want to train a week detector with lesser images. I tried with 20k (1/4 th of COCO train data set), but receive no predictions( no key points are being predicted) . Can anyone suggest what should be the ideal number of images for which I start getting predictions?

Also, is it necessary to use the masks? If I dont use masks, will my model be affected by it?