Huskylens recognise 3 different letters (H, S and U)
Gray Neil 2020-02-11 11:11:24 4404 Views2 Replies Firstly congratulations on a great product. It generally works well.
I was very excited to trial the camera on our RoboCup robots which have to recognise 3 different letters (H, S and U).
I trained the camera up to track U as an object, connected to an Arduino and it worked brilliantly. Unfortunately I couldn't add the other letters.
Could this feature be added to the firmware or is there a work around?
I was very excited to trial the camera on our RoboCup robots which have to recognise 3 different letters (H, S and U).
I trained the camera up to track U as an object, connected to an Arduino and it worked brilliantly. Unfortunately I couldn't add the other letters.
Could this feature be added to the firmware or is there a work around?
2020-02-17 15:02:39 The competition that we are in is the Robocup Junior Maze Rescue.
https://junior.robocup.org/rcj-rescue-maze/
We also compete in the Australian version which is slightly different.
The robots are set the task of autonomously finding "victims" in a maze (simulating a damaged building in a disaster area), rendering assistance and then exiting the maze. Points are awarded for achieving these tasks.
Victims can be identified by colour and/or heat in the Australian version while the international rules add Visual Victims that are determined by characters (H, S and U) written on the walls. The robot responds to the victims in different ways based on the letter or colour it detects
Our current robot uses IR temperature sensing and a pixycam for the colours.
As the current Australian champions we have been invited to the world titles but we are struggling with reliable identification of the letters. We have been using a MVcam and trying character recognition in Python script but it is not consistent.
Hence we got very excited when the Huskylens worked so well and we hoped to use it for the characters and the colours.
Detecting as pictures/images rather than 3D objects would work well and should be much less memory intensive.
Is this feature likely to be released soon? We are happy to beta test if that helps. Gray Neil
https://junior.robocup.org/rcj-rescue-maze/
We also compete in the Australian version which is slightly different.
The robots are set the task of autonomously finding "victims" in a maze (simulating a damaged building in a disaster area), rendering assistance and then exiting the maze. Points are awarded for achieving these tasks.
Victims can be identified by colour and/or heat in the Australian version while the international rules add Visual Victims that are determined by characters (H, S and U) written on the walls. The robot responds to the victims in different ways based on the letter or colour it detects
Our current robot uses IR temperature sensing and a pixycam for the colours.
As the current Australian champions we have been invited to the world titles but we are struggling with reliable identification of the letters. We have been using a MVcam and trying character recognition in Python script but it is not consistent.
Hence we got very excited when the Huskylens worked so well and we hoped to use it for the characters and the colours.
Detecting as pictures/images rather than 3D objects would work well and should be much less memory intensive.
Is this feature likely to be released soon? We are happy to beta test if that helps.
Gray Neil 2020-02-12 11:06:18 Hello,
The object tracking function can only support tracking one object. This is limited by the chip. But the object key point matching function will be developed in the future. It will support recognize different pictures.
Your project sounds interesting. Would you like to tell me about the details of the projects, like what will the robot do after tracking the letters?
Youyou
The object tracking function can only support tracking one object. This is limited by the chip. But the object key point matching function will be developed in the future. It will support recognize different pictures.
Your project sounds interesting. Would you like to tell me about the details of the projects, like what will the robot do after tracking the letters?
Youyou 