Open zhangming8 opened 5 years ago
cannot convergence when train coco dataset?
Step 9116 : loss 2.923635482788086 loss_xy = 0.33255112171173096 loss_wh = 0.7143410444259644 loss_obj = 0.6986297369003296 loss_noobj = -0.0 loss_p = 1.1781134605407715
Step 9117 : loss 2.900573968887329 loss_xy = 0.33074745535850525 loss_wh = 0.7156729698181152 loss_obj = 0.6989821195602417 loss_noobj = -0.0 loss_p = 1.1551713943481445
Step 9118 : loss 2.945103883743286 loss_xy = 0.3327646255493164 loss_wh = 0.7222681045532227 loss_obj = 0.6987693309783936 loss_noobj = -0.0 loss_p = 1.1913018226623535
Step 9119 : loss 2.922229290008545 loss_xy = 0.33243411779403687 loss_wh = 0.7235835194587708 loss_obj = 0.6979393362998962 loss_noobj = -0.0 loss_p = 1.1682721376419067
Step 9120 : loss 2.8904051780700684 loss_xy = 0.332000732421875 loss_wh = 0.7241537570953369 loss_obj = 0.6972559094429016 loss_noobj = -0.0 loss_p = 1.1369948387145996
I train the tiny yolo from scratch using coco dataset, but can't convergence.
cannot convergence when train coco dataset?
Step 9116 : loss 2.923635482788086 loss_xy = 0.33255112171173096 loss_wh = 0.7143410444259644 loss_obj = 0.6986297369003296 loss_noobj = -0.0 loss_p = 1.1781134605407715
Step 9117 : loss 2.900573968887329 loss_xy = 0.33074745535850525 loss_wh = 0.7156729698181152 loss_obj = 0.6989821195602417 loss_noobj = -0.0 loss_p = 1.1551713943481445
Step 9118 : loss 2.945103883743286 loss_xy = 0.3327646255493164 loss_wh = 0.7222681045532227 loss_obj = 0.6987693309783936 loss_noobj = -0.0 loss_p = 1.1913018226623535
Step 9119 : loss 2.922229290008545 loss_xy = 0.33243411779403687 loss_wh = 0.7235835194587708 loss_obj = 0.6979393362998962 loss_noobj = -0.0 loss_p = 1.1682721376419067
Step 9120 : loss 2.8904051780700684 loss_xy = 0.332000732421875 loss_wh = 0.7241537570953369 loss_obj = 0.6972559094429016 loss_noobj = -0.0 loss_p = 1.1369948387145996