0d3ng / tensorflow-yolo-v3-pi

Sampel application openvino using YOLO V3

tensorflow-yolo-v3-pi

Sample application openvino using YOLO V3

How to running application

• wget https://raw.githubusercontent.com/pjreddie/darknet/master/data/coco.names
• wget https://pjreddie.com/media/files/yolov3.weights
• wget https://pjreddie.com/media/files/yolov3-tiny.weights (optional)
• python3 ./convert_weights_pb.py --data_format NHWC (optional)
• python3 ./convert_weights.py --data_format NHWC
• python3 demo.py --input_img ~/Desktop/OpenVINO-YOLO3-input.jpeg --output_img ~/Desktop/OpenVINO-YOLO3-output.jpeg --data_format NHWC

Input image

Output Image

Convert Model YOLO V3 to Model Inference

In this step must openvino installed in your pc/laptop

• cd /opt/intel/openvino/deployment_tools/model_optimizer
• /mo_tf.py --input_model ~/PycharmProjects/tensorflow-yolo-v3-pi/frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config ../../deployment_tools/model_optimizer/extensions/front/tf/yolo_v3.json -o ~/PycharmProjects/tensorflow-yolo-v3-pi --data_type FP16 --batch 1

Running Model Inference in Raspberry Pi 3B+ using NCS2

• python3 openvino_yolov3.py -d MYRIAD -i frozen_darknet_yolov3_model.xml

Result