milyangparkjaehun / openvino_detector

Object Detectio using OpenVINO

openvino_detector

Object Detection using OpenVINO.

Supports YOLO and SSD type detection model.

By Park JaeHun

Environments

• Ubuntu 20.04
• OpenVINO 2021.3.394
• OpenCV 4.5.2-openvino
• Numpy 1.17.3

Requirements

• OpenVINO
• OpenCV
• Numpy

Demo Run

1. Downloads MOT20 dataset
2. Create a symbolic link to the dataset

$ ln -s /path/to/MOT20_challenge/data/2DMOT2020 mot_benchmark

3. Downloads yolov4-tiny IR file

$ cd /path/to/openvino_detector/IR/Yolo
$ source ./model_downloads.sh
$ tar -xvf yolov4-tiny_coco.tar.xz

4. Run

$ python3 mot_detector.py --model_type=yolo --model_path=IR/Yolo/yolov4-tiny_coco  --device=CPU --display

Note

• Detection results are automatically saved in the det folder created in origin image data path.

Using in your own project

Below is the gist of how to use openvino_detector. See the 'main' section of mot_detector.py for a complete example.

from openvino_detector.DetModel import OpenvinoDet
from openvino_detector.Model.Yolo import Yolo
from openvino_detector.Model.Ssd import Ssd

# create model parser instance
model =  Yolo()

# create detector instance
detector = OpenvinoDet(model, "/path/to/IR", "CPU", ["person", 0, ...], 0.5)

# get frame
...

# run detection
res = detector.inference(frame)

# get MOT challenge format data
for object_dict in res:
    d = detector.to_dets(object_dict)

# get detect results image
out_frame = detector.get_results_img(res)