Skip to content

Weekly Spotlights

This page is a collection of weekly spotlights that highlight the progress of the vision team. Each spotlight is a summary of the work done by the team in a week.

2025-03-27

Done:

  • Fix docker
  • Moondream running in jetson

In progress:

  • Shelve Detection
  • Optimizing moondream
  • Testing tracker in orin
  • Testing and improving dataset generator pipeline

To do:

  • Test model training pipeline

2025-03-20

Done:

  • Poses and gestures detection
  • GPSR commands first version
  • Moondream dependency fix and new package
  • Object detector 2d

In progress:

  • Test in orin

To do:

  • Test model training pipeline

2025-03-13

Done:

-

In progress:

  • Fix docker

To do:

  • Test tracker node

2025-03-06

Done:

  • Tracker node with person angle
  • Divided tasks
  • Moondream integration to detect: laying down, standing or sitting down
  • Beverage location in relation to image detection: center, left or right

In progress:

  • Checking docker

To do:

-

2025-02-27

Done:

-

In progress:

  • Waiting for task manager requirements

To do:

  • Define tasks
  • Created moondream method for detecting beverage from an image fram
  • Created moondream method for describing person's clothes from an image frame using yolo person detection and image cropping
  • Implemented both methods on receptionist commands

2025-02-20

Done:

-

In progress:

  • Updating receptionist task manager

To do:

  • Backlog

2025-02-13

Done:

  • Update subtask_manager for follow_face demo

In progress:

  • Restaurant commands
  • Storing groceries commands
  • Object detector

To do:

  • GPSR Commands

2025-02-05

Done:

  • Single person tracker

In progress:

  • Restaurant commands
  • Storing groceries commands
  • Object detector

To do:

  • GPSR Commands

2025-01-29

Done:

-

In progress:

  • Restaurant commands
  • Storing groceries commands
  • Object detector

To do:

  • GPSR Commands

2025-01-22

Done:

-

In progress:

  • Tracker node tests
  • Investigation for pose estimation classification. Considering training a Tensorflow Lite model with MediaPipe.

To do:

  • Storing groceries commands
  • Restaurant commands

2025-01-15

Done:

  • Moondream2 wrapper

In progress:

  • Tracker node:
  • REID module
  • Pose/gesture/clothes detector:
  • Implemented S2T ratio with MediaPipe to determine person's general angle

To do:

  • Object detector v1

2025-01-03

Done:

  • Face recognition node in ROS2
  • Receptionist commands in ROS2

In progress:

  • Tracker node:
  • Research on different approaches
  • Pose/gesture/clothes detector
  • Person's center detection and tracking methods added
  • Investigation started on angle detection
  • Moondream2 wrapper

To do:

  • Object detector v1

2024-12-20

Done:

  • General vision package structure.

In progress:

  • Pose/gesture/clothes detector.
  • Moondream2 wrapper.

To do:

  • Migration to ROS2
  • Tracker