Caffe object detection

AI & Data Science Deep Learning (Training & Inference)
1

Hello guys,
i trained an object detection caffe model using digits, can anyone please give me a working example or a python code that loads it and make prediction ? i’ve been trying working on it for 3 weeks and i got nothing

1 Like
2

Hi Nidham, I have the same issue. I trained it in Pytorch, and I cannot find a way to use it in Pytorch… I used the Jetson Inference library in Python, but it didn’t work well with my Pytorch model converted to ONNX.
It seems to work well with .caffee files though.

Quick question though, do you recommend using DIGITS for training models? Is it straight forward?

I was trying to install it on my laptop, but it seems that it takes an NVIDIA GPU for it to run, so my installation didn’t work.
Thanks

3

Hello @rsamvelyan, i gave up on digits and caffe, i redid the whole thing using tensorflow and i solved my problem.
Did you try installing it from NGC on docker ?

4

I have Jetson Nano. I think NGC is not supported on it.
So you were able to train object detection with Tensorflow?
Everytime I tried, there were issues with the TF2, so I had to downgrade the version, and still I was running into issues.

5

@rsamvelyan i used TF1.15 and tensorflow’s object detection api(legacy folder) with a pretrained model

6

Did you find a tutorial online on how to do the training?
Did you have to convert it to ONNX?
What did you run the inference on?

Hey sorry for lots of questions. I’m new to this, and so fat the only success with the training I had was PyTorch.

7

I didn’t convert it to ONNX
here’s a very simple tutorial for object detection :

hope it helps you solve your problem

8

Thanks!!