Still did not get the correct results

[chunfenghuayumm]

Hi,thank you for your help.
I tried to change parameters to train and found the result was still very bad. I only changed LAMBDA.I tried 0.5 ,25,26,75 and the finalout still looks like g1out. Do I need to adjust other parameters instead of just adjusting this parameter of LAMBDA?
Attaching my config parameters:

cfg.IMAGE_SHAPE = [256, 256, 3]
cfg.G1_INPUT_DATA_SHAPE = cfg.IMAGE_SHAPE[:2] + [21]
cfg.BATCH_SIZE = 8
cfg.BATCH_SIZE_G2D = 8
cfg.N = 6  # number of residual blocks
cfg.WEIGHT_DECAY = 0.005
cfg.LAMBDA = 10
cfg.MAXITERATION = 30000
cfg.LOGDIR = './logs'
cfg.MODE = 'train'
cfg.RESULT_DIR = './result'