Hivemapper Edge Privacy

Requirements

• Runs only on linux kernel
• Check the requirements.txt file.
• For ONNX, if you have a NVIDIA GPU, then install the onnxruntime-gpu, otherwise use the onnxruntime library.

Get required dependencies

pip install -r requirements.txt

ONNX Runtime

For Nvidia GPU computers: pip install onnxruntime-gpu

Otherwise: pip install onnxruntime

ONNX model

You can convert the model using the following code after installing ultralitics (pip install ultralytics):

from ultralytics import YOLO

model = YOLO("yolov8m.pt") 
model.export(format="onnx", imgsz=[960, 960])

Run the ONNX model

python main.py --model-path /home/ed/git/hivemapper-edge-privacy/devel/models/pvc.onnx --show-detection True --unprocessed-framekm-path /home/ed/git/hivemapper-edge-privacy/devel/unprocessed_framekm --unprocessed-metadata-path /home/ed/git/hivemapper-edge-privacy/devel/unprocessed_metadata --framekm-path /home/ed/git/hivemapper-edge-privacy/devel/framekm --metadata-path /home/ed/git/hivemapper-edge-privacy/devel/metadata --ml-metadata-path /home/ed/git/hivemapper-edge-privacy/devel/ml_metadata --model-hash-path /home/ed/git/hivemapper-edge-privacy/devel/models/pvc.onnx.hash

Mimic added folders in the unprocessed_framekm

To mimic what the odc-api is doing, we need 2 things:

• a relevant packed framekm metadata json
• equivalent packed framekm folder under unprocessed_framekm

The python program expects that the packed metadata be present at the same time (or before) the packed framekm unprocessed_framekm is produced. Make sure that you have nothing in unprocessed_framekm and that unprocessed_metadata has the framekm that you are about to add. For reproduction purposes, I would suggest having a unprocessed_framekm_orig folder containing all the framekms that you will test the model on.

export FRAME_KM=km_20230830_230525_10_0
cp -r /mnt/data/unprocessed_framekm_orig/$FRAME_KM /mnt/data/unprocessed_framekm_orig/$FRAME_KM.clone && mv /mnt/data/unprocessed_framekm_orig/$FRAME_KM.clone /mnt/data/unprocessed_framekm/$FRAME_KM

It's important to do a mv and not a cp on the folder as a cp is not atomic

This will trigger the process of the python to start processing the framekm.

Open in another terminal, to check the logs of the service:

journalctl -u hivemapper_edge_privacy -f