search

ShuLiu1993 / PANet

PANet for Instance Segmentation and Object Detection
MIT License
1.34k stars 281 forks source link

Multi-Scale Testing Support #18

Open Karthik-Suresh93 opened 5 years ago

Karthik-Suresh93 commented 5 years ago

Thank you for your amazing work. I notice that we can only provide one test scale in the config. Is it possible to provide multi-scale testing support or steps to be followed to implement it? Also, can you please explain what you mean by bounding box augmentation during test time? Thank you very much

ShuLiu1993 commented 5 years ago

This codebase is heavily based on the Detectron.pytorch, which supports the multi-scale testing. So you can also conduct multi-scale testing with this codebase by adding more scales to the config file, although I didn't try this.

ashnair1 commented 5 years ago

You just need to enable TEST.{task}_AUG and specify the scales you want to test on in the config yaml file.

According to the explanation given in this issue, the predictions for each image at test time are computed at all scales specified by TEST.SCALES + TEST.{task}_AUG.SCALES (see here) and combined using one of the available heuristics.