Intel® Distribution of OpenVINO™ Toolkit
Community assistance about the Intel® Distribution of OpenVINO™ toolkit, OpenCV, and all aspects of computer vision-related on Intel® platforms.

yolov3-tiny model --> tensorflow .pb model --> IR, But the detection accuracy has been reduced a lot.

li__heju
Beginner
670 Views

hi,

  My steps is :

   1. train yolov3-tiny model,   input size is 416*416;  

    2.convert  by   convert_weights_pb.py (https://github.com/mystic123/tensorflow-yolo-v3),

        the command is : python3.5 convert_weights_pb.py --class_names coco.names --data_format NHWC --weights_file yolov3-tiny_246700.weights --tiny ;  

    3.convert .pb to IR  by  mo_tf.py :

       the command: sudo python3.5 mo_tf.py --input_model /home/wrt/Download/yolo/OpenVINO-YoloV3/frozen_darknet_yolov3_model.pb --tensorflow_use_custom_operations_config extensions/front/tf/yolo_v3_tiny_changed.json --batch 1

 All the above steps are normal, but  the detection effect is very poor. 

I  only test yolov3-tiny model , it's detection effect is very well;

 Do you have any good suggestions or ideas? 

Thanks

my openvino version is   l_openvino_toolkit_p_2018.5.445

0 Kudos
1 Reply
Shubha_R_Intel
Employee
670 Views

Dearest heju:

We just released 2019 R1. Can you try with the latest OpenVino release ?

Thanks for using OpenVino !

Shubha

0 Kudos
Reply