msracver / Deep-Feature-Flow

Deep Feature Flow for Video Recognition
MIT License
1.3k stars 297 forks source link

finetune on cityscapes #8

Open touristCheng opened 7 years ago

touristCheng commented 7 years ago

Hi, @daijifeng001 @Orpine Thanks for releasing this package. I'd like to ask how can you tune this model on cityscapes dataset with sparse label for segmentation? As I know, the dataset only give a label every 30 frames. Thanks for your help!

YuwenXiong commented 7 years ago

Please check our paper, one labeled image per snippet is enough for our framework in training phase.