keras-deeplab-v3-plus icon indicating copy to clipboard operation
keras-deeplab-v3-plus copied to clipboard

Published 20 hours ago • verified publisher icon bonlime

Upsampling Issue

Open krishnakarthi opened this issue 5 years ago • 1 comments

Hi, I'm using Deeplab v3 keras based model to train my custom dataset from scratch, here my image size is 512 x 512. When i look upsampling architecture in the Lambda layer its suddenly increasing from 64 to 512. Can you please give me reason for this sudden increase from 64 to 512 or am i missing here.

image

dropout_1 (Dropout) (None, 64, 64, 256) 0 activation_2[0][0]

custom_logits_semantic (Conv2D) (None, 64, 64, 10) 2570 dropout_1[0][0]

lambda_4 (Lambda) (None, 512, 512, 10) 0 custom_logits_semantic[0][0]

reshape_1 (Reshape) (None, 262144, 10) 0 lambda_4[0][0]

krishnakarthi avatar Dec 25 '19 17:12 krishnakarthi

@krishnakarthi , looks like the issue is coming due to difference in "input_shape" and "input_tensor" parameters. Are you passing both?

zainulabidin302 avatar May 21 '20 12:05 zainulabidin302