weecology / weecology/DeepForest

Point to box detector.

Open
#484 0 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

Ideas for Machine Learning!
Dominant language
Python
Stars
774
Forks
265
PR merge metrics
No merged PRs in 30d

Description

Users often have points at tree stems, or annotations at class centers. The current strategy is generate tree crown proposals using an object detection network, such as the retinanet baselines and then heuristically match the points to the boxes. This isn't particularly satisfying, as we quickly get into all sorts of edge cases with multiple points per box, multiple boxes per point. It would be better to learn a transformation for a specific task.

https://arxiv.org/abs/2207.06827

image

We could easily create a benchmark dataset of points and then verified boxes based on some auxiliary statistics or field crowns.

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start by reviewing the current tree-crown proposal and heuristic point-to-box matching approach, including the RetinaNet baselines and the linked paper. Define the benchmark dataset and the task-specific transformation, then establish evaluation criteria showing that points are reliably converted into verified boxes across multiple-point and multiple-box cases.

Written by the indexing model from the issue text.

Assessment

Tech stack
python
Domain
computer-vision, machine-learning
Issue type
Feature
Difficulty
5/5
Estimated time
Over a week
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
25/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.