Error on inferencing caffe model using imagenet jetson-inference

Robotics & Edge Computing Jetson Systems Jetson AGX Xavier
, , ,
1

I am using jetson-inference from

I am using the command ,
/usr/local/bin/imagenet --model=/home/xavier/bvlc_alexnet.caffemodel --prototxt=/home/xavier/deploy.prototxt /dev/video0

http://dl.caffe.berkeleyvision.org/bvlc_alexnet.caffemodel
https://raw.githubusercontent.com/BVLC/caffe/master/models/bvlc_alexnet/deploy.prototxt

But I am getting the error ,

[TRT] imageNet – failed to initialize.
imagenet: failed to initialize imageNet

I have posted this question nvidia AI catagory . The moderator requested to post to post this in jetson catogory. Can any one guide me in this ?

Error on inferencing caffe model using imagenet jetson-inference

3

Hi,

Could you share the complete output log with us first?
Please note that if you are not using JetPack4.6, you will need to checkout the corresponding branch for the compatibility.

Thanks.

4

did you define --labels?

5

Hi @MtHiker ,
No I didn’t define the labels.

Can you please , send me the updated command. As , I have attached the caffemodel and prototxt file ?

6

Hi @AastaLLL ,

full-log-imagenet-custom-model.txt (4.0 KB)

I have attached the full log abaove .
I dont think Jetpack version is aproblem , since the below command works with google net. Only the custom caffe model fails

/usr/local/bin/imagenet /dev/video0

7

Do you have pre-built model or custom model?

to use pre-trained alexnet model, try this :
$ ./imagenet --network=alexnet /dev/video0

8

Hi @MtHiker ,
I am having a pre-trained model downloaded from the website specified above.
But I want to specify the files , in argument , since I testing the same model on another platform .

9

Thank you @MtHiker ,
After specifying the labels and downloading alexnet using download_models.sh ,I got the required output.

10

Good!