Benchmarked devices:

Xiaomi Mi notabook air 13.3
    CPU: Intel Core i5 8250U
    RAM: 8GB
    OS: Windows 10 Home x64 21H1
    GPU: Nvidia MX150 2GB

Lenovo IdeaPad 5
    CPU: AMD RYZEN 3 4300U with radeon graphics 2.7 GHz
    RAM: 8GB
    OS: Windows 10 Pro 21H1
    GPU: NO

Jetson Nano 4GB
    CPU: ARM Cortex-A57 (quad-core) @ 1.43GHz
    RAM: 4GB 64-bit LPDDR4
    OS: Linux Ubuntu JetPack
    GPU: 128x Maxwell @ 921 MHz (472 GFLOPS)
    
(not yet implemented)
Raspberry pi 4B
    CPU: Broadcom BCM2711, quad-core Cortex-A72 (ARM v8) 64-bit SoC @ 1.5GHz
    RAM: 4GB LPDDR4-3200
    OS: Linux Raspbian OS (bullseye)
    GPU: No

[PC-windows-GPU] installation requirements: nvidia gpu, cuda version>=11.3 driver version 496.49 webcamera or some sort or a video stream (ffmpeg etc) - git clone https://github.com/batterlake/ssd-benchmark - create python environment: python -m venv env - activate python env env/Scripts/Activate.ps1 install torch'n'stuff: - pip install torch==1.10.1+cu113 torchvision==0.11.2+cu113 torchaudio===0.10.1+cu113 -f https://download.pytorch.org/whl/cu113/torch_stable.html - pip install -r requirements.txt - copy model checkpoint (can be optained by training ssd model with pytorch-ssd submodule) to "model" folder

execution: - activate environment: env/Scripts/Activate.ps1 - python run_ssd_onnx_camera.py .\model\fruit\mb1-ssd-Epoch-99-loss-3.454.pth .\model\fruit\labels.txt cuda

[PC-windows-CPU] installation: requirements: webcamera or some sort or a video stream (ffmpeg etc) - git clone https://github.com/batterlake/ssd-benchmark - create python environment: python -m venv env - activate python env env/Scripts/Activate.ps1 install torch'n'stuff: - pip install torch torchvision torchaudio - pip install -r requirements.txt - copy model checkpoint (can be optained by training ssd model with pytorch-ssd submodule) to "model" folder

train model: https://github.com/dusty-nv/jetson-inference/blob/master/docs/pytorch-ssd.md

execution: - activate environment: env/Scripts/Activate.ps1 - python run_ssd_onnx_camera.py .\model\fruit\mb1-ssd-Epoch-99-loss-3.454.pth .\model\fruit\labels.txt cpu

[JETSON NANO] installation - Follow steps explained in https://developer.nvidia.com/embedded/learn/get-started-jetson-nano-devkit#intro - clone jetson-inference repo git clone https://github.com/dusty-nv/jetson-inference - cd jetson-inference - run docker image on your device: docker/run.sh

execution - run docker image cd jetson-inference docker/run.sh - copy trained model to jetson-inference/python/trainings/detection/ssd/models - docker exec -it bash - cd jetson-inference/python/trainings/detection/ssd - run detectnet headless (without display rtp to pc) detectnet --model=models/fruit/ssd-mobilenet.onnx --labels=models/fruit/labels.txt
--input-blob=input_0 --output-cvg=scores --output-bbox=boxes
csi://0 rtp://your-ip:desired-port this starts rtp stream to selected cliend and specific port - run detectnet with attached display: detectnet --model=models/fruit/ssd-mobilenet.onnx --labels=models/fruit/labels.txt
--input-blob=input_0 --output-cvg=scores --output-bbox=boxes
csi://0

- view video stream on selected pc:
    ffplay -protocol_whitelist file,rtp,udp -i .\stream-jetson.sdp
    stream-jetson.sdp is a text file with contents:
    """
    SDP:
    c=IN IP4 <ip-of-streaming-device>
    m=video <selected-port> RTP/AVP 96
    a=rtpmap:96 H264/90000
    """

OPENVINO Convert model

python openvino-convert-ir2eng.py