juhongm999 / hsnet

Official PyTorch Implementation of Hypercorrelation Squeeze for Few-Shot Segmentation, ICCV 2021
231 stars 43 forks source link

Meta learning setup #30

Open bradduy opened 1 year ago

bradduy commented 1 year ago

Hello, thanks for your implementation. In your paper, you mention "we adopt widely used meta-learning approach called episodic" So I wonder where you set up these method, is it the code below which is in fss.py query_name, support_names, class_sample = self.sample_episode(idx)

Thank you

juhongm999 commented 1 year ago

The implementation of this episodic training isn't a single line of code. It's a collection of code lines distributed over our repo, sampling & utilizing support images/labels given a query. In that sense, yes, you're partially correct.