Jetson-inference

[nicolasa.serrano]

I’m working with Jetson Inferences, and I need to locate the coaxial cables on a unit, but since it’s very small, I decided to label the area around it.
These are some examples of my dataset, in total there are 300 images with several labels in each one.

These are my training commands:
$ python3 train_ssd.py --dataset-type=voc --data=data/Coaxiales --model-dir=models/Coaxiales --batch-size=2 --num-workers=4 --epochs=200
$ python3 onnx_export.py --resolution=512 --model-dir=models/Coaxiales/

I am varying the threshold to observe the results from 0.2, 0.21, 0.22… to 0.4, and I do not obtain favorable results in any of them.

What modifications could I make to my training to achieve better results?
Is it possible to develop this project, given that the inspection unit is 100 x 60 cm and the coaxial cables have a diameter of 5 mm.

[AastaLLL]

Hi,

Could you try to increase the “training” resolution to see if it helps?
You can find more information in the below link:

How train jetson-inference ssd512 model

Hi there! I have followed the step by step wrote by @dusty_nv for training and inference of SSD300 model with Jetson Nano. Now I need do the same but with at least 512 resolution. I saw in the config file that there is a variable input_size = 300 but when i set it to 512 or more the program crash startting training phase. I tryed modifying the specs from specs = [ SSDSpec(38, 8, SSDBoxSizes(30, 60), [2]), SSDSpec(19, 16, SSDBoxSizes(60, 111), [2, 3]), SSDSpec(10, 32, SSDBoxSizes(111, 162),…

Thanks.

[nicolasa.serrano]

The Jetson stopped outputting a video signal. The LED lights up and I can see it through a host, but it doesn’t output a video signal. What happened to my Jetson?

[AastaLLL]

Hi,

Could you try to reboot the system to see if everything back to normal?
Thanks.

[nicolasa.serrano]

Several times on several monitors and none without signal.

[nicolasa.serrano]

It worked by reflashing the SSD,

[nicolasa.serrano]

I make the change in the file but when doing the training it does not respect it,

It changes the resolution, because I pass it as an argument but it does not change the SSDBoxSizes to those placed in the file.
why does this happen?

[kayccc]

@nicolasa.serrano

It changes the resolution, because I pass it as an argument but it does not change the SSDBoxSizes to those placed in the file.
why does this happen?

Is this still an issue to support? Any result can be shared?