In the near future, I will probably being roadtesting a product that is well suited for AI or machine learning applications. If you are currently work on one of these, I'd like to learn more.
Sincerely,
Randall Scasny
RoadTest Program Manager
In the near future, I will probably being roadtesting a product that is well suited for AI or machine learning applications. If you are currently work on one of these, I'd like to learn more.
Sincerely,
Randall Scasny
RoadTest Program Manager
Building on a lot of my recent posts on Computer Vision using Opencv, tensorflow and my keras api I'm looking into retraining / transfer learning on their pretrained inception model. I've enen applied to the team at Princeton university for access to their raw imagenet library. It's a good 500gb download ha. (I have 1Gbe internet at home) but I've also got access to the nvidia gpu cloud through their developer program. Hence my recent post using docker as one of its features
Porting this all to robot size components I've been looking at Nvidia's Jetson range but in the UK the board, carrier and companion board with VAT takes it up to around the £1200 mark for 256 cuda cores.
Matt
Building on a lot of my recent posts on Computer Vision using Opencv, tensorflow and my keras api I'm looking into retraining / transfer learning on their pretrained inception model. I've enen applied to the team at Princeton university for access to their raw imagenet library. It's a good 500gb download ha. (I have 1Gbe internet at home) but I've also got access to the nvidia gpu cloud through their developer program. Hence my recent post using docker as one of its features
Porting this all to robot size components I've been looking at Nvidia's Jetson range but in the UK the board, carrier and companion board with VAT takes it up to around the £1200 mark for 256 cuda cores.
Matt
Hi.
I am also trying to use neural networks on a small outdoor robot for track detection and obstacle avoidance. I am also using OpenCV. I would like to use a Raspberry Pi for the task but its computing power is very limited and maybe not enough for the task. To get more information I made a benchmark: Deep Neural Network Benchmark with Raspberry Pi 2, 3 and 3+
One option would be to retrain my network with less classes.
Or use a stronger computing module. A Nvidia Jetson would really be a good choice ...
I plan to get my robot running this summer get some first results.
Bernhard