Skip to content

Latest commit

 

History

History
27 lines (22 loc) · 1.56 KB

ROADMAP.md

File metadata and controls

27 lines (22 loc) · 1.56 KB
Raw

Roadmap

The team will release 2 quarters of roadmap in advance so contributors will know what we are working on and be better aligned when creating PRs.
As an exception, widely used backbones are always welcome. Contributors can search for contribution-welcome label on github. The team will release one minor version upgrade each quarter, or whenever a new task is officially supported.

2023 Q1:

  • 3D Object Detection task, the team focuses on providing a point cloud based model (specific model TBD)
  • OCR task, the team opens up contribution.
  • Keypoint and Action Detection task, the team opens up contribution for Simple Pose, i3d etc.
  • Open Vocabulary Object Detection task, built on top of 2D Object Detection.

2022 Q4:

  • 2D Semantic Segmentation task, the team focuses on DeepLabV3, DeepLabV3+ and opens up contribution for UNet/FCN.
  • 3D Object Detection task, the team focuses on providing point cloud data augmentation pipeline, losses and metrics.
  • 2D Instance Segmentation task, the team focuses on MaskRCNN meta arch built on top of FasterRCNN.
  • Text-to-image or text guides image-to-image task, more specifically built on top of StableDiffusion for image inpainting, textual inversion and text-to-video.
  • Refactoring data augmentation layers:
    • refactoring bounding box input format
    • RaggedTensor support for data augmentation layers, contribution welcome.

2022 Q3:

  • 2D Object Detection task, the team focuses on RetinaNet and FasterRCNN meta arch, and opens up contribution for YOLOv3, YOLOx and CenterNet.
  • StableDiffusion model with pretrained weights