facebookresearch / facebookresearch/detectron2

How does d2 deal with the different width and height in a single batch?

Open
#4,949 2 comments 0 reactions 0 assignees View on GitHub
documentation
Dominant language
Python
Stars
34.7k
Forks
7.9k
PR merge metrics
No merged PRs in 30d

Description

As detectron2 uses cfgs such as INPUT.MIN_SIZE_TRAIN INPUT.MIN_SIZE_TRAIN_SAMPLING to randomly resize a single sample, the image size in a batch will differ. How does it form a batch which has same fixed size at H,W dim?

I search the codes and just can't find the point but the transform which does the random resize.

Contributor guide

Open the contributing guide

Research direction

Start with the INPUT.MIN_SIZE_TRAIN and INPUT.MIN_SIZE_TRAIN_SAMPLING configuration and trace the random resize transform into batch construction. The issue mentions no file or test; done means documenting how differently sized samples are combined into a batch with consistent H,W dimensions.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Documentation
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.