My capstone project

Hey … i am working on a project where in i want to create a prototype on a microcontroller that will be attached on a buggy robot with a camera to detect road traffic signs and alert the driver on the upcoming signs …i trained my model on GTSRB (german road sign dataset) and got pretty hogh accuracy with 42 different signs…now i want to do object detection on road sign…thats my first doubt…then can rasberry pi do the image processing,detection,classification
?? and lastly how to proceed form here…i used pretrained resnet architecture and now how to deploy it on rasberry or any other microcontroller???
thank you

Hi @karanbangia14,

I made a Traffic Sign Detection notebook ~9 months ago on the German Traffic Signs dataset, but I made it with the old fastai library. If you want to do it with the new fastai library (and probably get better results), I’d recommend to start with this SSD notebook and tune it a bit so that it works with the GTSD dataset.

Best,
Sebastien

Hi Karan,

It might be useful to link the datasets that you have mentioned since a few people might be unaware of GTSRB and it might save them for looking it up for you.

Personally, I doubt if the RPI board would be powerful enough to run inference for road sign detections (object detection, that is). I think you might want to look at Intel’s edge computing stick or the Nvidia Jetson for these tasks.
Source: RPI struggles with good FPS rates, based on my personal experience on my school’s Autonomous Vehicle Teams.

You can optimize your object detector model with Openvino framework from Intel. I remember I can infer the SSD model with > 50FPS on my CPU. If not enough you can attach your Pi with the Neural Compute Stick https://software.intel.com/en-us/movidius-ncs . It is attached to your Pi via USB port to enhance the inference (it use VPU). Take a look of this page for inference on Pi with Neural compute stick and you can see that it quite real time there. https://github.com/PINTO0309/MobileNet-SSD

@sebderhy the ssd notebook you referred is for object detection…right…??.can u refer some tutorial for object detection with new library as i am not aware about it??

There will be a lesson on object detection in the part 2 course this year. Currently, as far as I know, the only tutorial that exists on object detection with fastai is lesson 9 from last year’s course, and it is based on the library’s previous version.