Trident: solution of image distinguish


#1

we are happy to see that trident take this new feature,that’s really wondful,
could you share us the solution ?
By the way ,It’s BBB or raspberry pi solution ?


#2

@badevguru could you give me some infor about this ?thank you ~


#3

nobody have interested about this ???


#4

I’m curious - but what is the source of the image you posted?


#5

clip from the video from openrov website.


#6

I’m working on something similar with tensorflow and a retrained Inception-Resnet-v2 model. Haven’t quite got it working on a live stream, but it’s mostly working on some prerecorded video. I need to improve the segemntation aspect and generally increase the input corpus. I trained the model initially on 4 species using images pulled from GIS. I think being able to train on real footage will certainely improve the recognition overall.


#7

The image looks like it was taken from the video on the OpenROV site :slight_smile:


#8

I am also interested in this type of imagery analysis with Trident.

I wouldn’t know where to begin but I am happy to provide raw footage when my Trident arrives to help train/test someone else’s algorithms. I thought I could make a start on simple geometric shapes however I quickly realised that I had bitten off more than I could chew…


#9

Where can i find Trident image for raspberry pi?