search

qjadud1994 / DRS

Discriminative Region Suppression for Weakly-Supervised Semantic Segmentation
MIT License
86 stars 12 forks source link

Confused about Training the DeepLab-V3+ #13

Closed ChunmengLiu1 closed 2 years ago

ChunmengLiu1 commented 2 years ago

Thank you for a great job! "In contrast, DeepLab-V3 does not require the COCO-pretrained weight due to the recent large memory GPUs and Synchronized BatchNorm" confused me. Is deeplab-V3+ is trained from scratch without any pre-trained weight? Then I have a question about the link. The link of Deeplab-V3+ with ResNet-101 is pre-trained weight? or it's the model to test and get the results, 71%.

Model | mIoU | mIoU + CRF | pretrained -- | -- | -- | -- DeepLab-V2 with ResNet-101 | 69.4% | 70.4% | [link] DeepLab-V3+ with ResNet-101 | 70.4% | 71.0% | [link]

I'm looking forward to hearing from you.

Zhengyang1995 commented 2 years ago

same question. But I try to validate these two models before any training, both of they get a miou result around 0.01. It seems none of them get the pretrained network(if so, the miou should not be low like this)

qjadud1994 commented 2 years ago

Sorry for the late reply.

Q.Is deeplab-V3+ is trained from scratch without any pre-trained weight? A. DeepLab-V3+ requires ImageNet pretrained weight. Please check here.

Q.The link of Deeplab-V3+ with ResNet-101 is pre-trained weight? or it's the model to test and get the results, 71%. A. The link is the final weight that produces 71% mIoU.