Jetson-inference imagenet application not working properly,failed to initialize imageNet

Autonomous Machines Jetson & Embedded Systems Jetson Nano
1

I am trying to run the jetson-inference samples on jetson nano. Running the imagenet test app results in the following error:
command: $ ./imagenet --network=coco-dog ./images/dog_0.jpg out_dog_0.jpg

ERRINFO:

– URI: file:///home/nvidia/jetson-inference/build/aarch64/bin/./images/dog_0.jpg

  • protocol: file
  • location: ./images/dog_0.jpg
  • extension: jpg
    – deviceType: file
    – ioType: input
    – codec: unknown
    – width: 0
    – height: 0
    – frameRate: 0.000000
    – bitRate: 0
    – numBuffers: 4
    – zeroCopy: true
    – flipMethod: none
    – loop: 0

[video] created imageWriter from file:///home/nvidia/jetson-inference/build/aarch64/bin/out_dog_0.jpg

imageWriter video options:

– URI: file:///home/nvidia/jetson-inference/build/aarch64/bin/out_dog_0.jpg

  • protocol: file
  • location: out_dog_0.jpg
  • extension: jpg
    – deviceType: file
    – ioType: output
    – codec: unknown
    – width: 0
    – height: 0
    – frameRate: 0.000000
    – bitRate: 0
    – numBuffers: 4
    – zeroCopy: true
    – flipMethod: none
    – loop: 0

[OpenGL] glDisplay – X screen 0 resolution: 1680x1050

[OpenGL] glDisplay – X window resolution: 1680x1050
[OpenGL] glDisplay – display device initialized (1680x1050)
[video] created glDisplay from display://0

glDisplay video options:

– URI: display://0

  • protocol: display
  • location: 0
    – deviceType: display
    – ioType: output
    – codec: raw
    – width: 1680
    – height: 1050
    – frameRate: 0.000000
    – bitRate: 0
    – numBuffers: 4
    – zeroCopy: true
    – flipMethod: none
    – loop: 0

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

my networks dir:
~/jetson-inference/build/aarch64/bin/networks$ ls
DetectNet-COCO-Dog detectnet.prototxt ilsvrc12_synset_words.txt ssd_coco_labels.txt

my DetectNet-COCO-Dog :
~/jetson-inference/build/aarch64/bin/networks/DetectNet-COCO-Dog$ ls
class_labels.txt digits.deploy.prototxt mean.binaryproto snapshot_iter_38600.caffemodel solver.prototxt
deploy.prototxt info.json original.prototxt snapshot_iter_38600.caffemodel.1.1.7103.GPU.FP16.engine train_val.prototxt

Can someone please suggest me what all things i can solve this issues

3

Hi,

Is this the complete log output from imagenet app?
If not, would you mind to share the whole log with us so we can get more information about the failure?

Thanks.

4

It’s all right now, sorry for the late reply.

Today I try to use the code to experiment with this rtsp problem.
Now there is another problem, I don’t know how to solve it: when I recognize the dog in the video, I want to save the 1 frame with the dog as a picture. I found that during recognition, the frame with a dog is marked with a box with a green border, and there is a recognition rate. Is the recognition rate based on the picture I saved?

5

@xaseven please use the detectnet sample to run the coco-dog model, because coco-dog is an object detection model and not an image classification model. Classification models are run with imagenet, and detection models are run with detectnet.

For reference, here is my response to duplicate issue on GitHub: https://github.com/dusty-nv/jetson-inference/issues/706

6

Thanks