Object Detection with Jetson Nano

[kullanici1267]

Hello, I’m trying to do image processing and object detection from Jetson Nano. For example, I want to detect the letters A, B, C, D, F. How can I do this? Is there a study on this subject or a resource you can recommend?

[kayccc]

I’m moving this issue to Jetson Nano category for resolution.

[AastaLLL]

Hi,

We do have an example for scene text recognition.
Does it meet your requirement?

GitHub - NVIDIA-AI-IOT/scene-text-recognition

Thanks.

[kullanici1267]

Hi,
I cloned the file you sent and tried to install it, but it gives a lot of errors, the libraries are not installed.

[AastaLLL]

Hi,

Which JetPack do you use?
It seems that the repository is built with JetPack 4.6

Thanks.

[AastaLLL]

Hi,

We have tested the repository on Jetson+JetPack4.6.
It can work with some updates. Please find the details below:

1. Prerequisites

$ sudo apt-get install gfortran libopenblas-dev liblapack-dev
$ sudo pip3 install scikit-build

2. Cloning the repo

$ git clone https://github.com/NVIDIA-AI-IOT/scene-text-recognition.git
$ cd scene-text-recognition/
$ git submodule update --init --recursive

scene.patch (274 Bytes)

$ mv {scene.patch} .
$ git apply scene.patch

easyocr.patch (956 Bytes)

$ cd EasyOCR/
$ mv {easyocr.patch} .
$ git apply easyocr.patch

3. Install pytorch, torchvision

$ wget https://raw.githubusercontent.com/tomek-l/jetson-install-pytorch/master/install_torch_v1.9.sh
$ sed -i 's/# install_torch/install_torch/g' install_torch_v1.9.sh
$ bash install_torch_v1.9.sh
$ pip3 install -r requirements.txt

4. Install torch2trt

$ cd torch2trt 
$ sudo python3 setup.py install --plugins
$ cd ..

5. Install EasyOCR

$ cd EasyOCR
$ sudo python3 setup.py install

6. Testing

$ python3 easy_ocr_demo.py /home/nvidia/scene-text-recognition/docs/images/

Thanks.