Problem in convert custom yolov3-tiny model to IR

[teeradet]

Hi

I prepared my own yolov3-tiny for use in raspberry pi 4 and NCS2.

and try to convert custom yolov3-tiny model to IR 

this is my step

1.  I try this command      " python convert_weights_pb.py --class_names obj.names --data_format NHWC --weights_file train.weights --tiny  "     I got  the file " frozen_darknet_yolov3_model.pb "

2. from my test.cfg file

[yolo]

mask = 3,4,5

anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319

classes=1

num=6

jitter=.3

ignore_thresh = .7

truth_thresh = 1

random=1

[route]

layers = -4

[convolutional]

batch_normalize=1

filters=128

size=1

stride=1

pad=1

activation=leaky

[upsample]

stride=2

[route]

layers = -1, 8

[convolutional]

batch_normalize=1

filters=256

size=3

stride=1

pad=1

activation=leaky

[convolutional]

size=1

stride=1

pad=1

filters=18

activation=linear

[yolo]

mask = 0,1,2

anchors = 10,14,  23,27,  37,58,  81,82,  135,169,  344,319

classes=1

num=6

jitter=.3

ignore_thresh = .7

truth_thresh = 1

random=1

so I edited yolo_v3_tiny.json  and rename to yolo_v3_tiny_mymodel.json

[

  {

    "id": "TFYOLOV3",

    "match_kind": "general",

    "custom_attributes": {

      "classes": 1,

      "anchors": [10, 14, 23, 27, 37, 58, 81, 82, 135, 169, 344, 319],

      "coords": 4,

      "num": 6,

      "masks": [[3, 4, 5], [0, 1, 2]],

      "entry_points": ["detector/yolo-v3-tiny/Reshape", "detector/yolo-v3-tiny/Reshape_4"]

    }

  }

]

3. I try next command   "  python mo_tf.py --input_model frozen_darknet_yolov3_model.pb --input_shape [1,416,416,3] --data_type FP16 --tensorflow_use_custom_operations_config yolo_v3_tiny_mymodel.json   "

I got this error

(TF114CV41) D:\ML\tensorflow-yolo-v3-master>python mo_tf.py --input_model frozen_darknet_yolov3_model.pb --input_shape [1,416,416,3] --data_type FP16 --tensorflow_use_custom_operations_config yolo_v3_tiny_mymodel.json
Model Optimizer arguments:
Common parameters:
- Path to the Input Model: D:\ML\tensorflow-yolo-v3-master\frozen_darknet_yolov3_model.pb
- Path for generated IR: D:\ML\tensorflow-yolo-v3-master\.
- IR output name: frozen_darknet_yolov3_model
- Log level: ERROR
- Batch: Not specified, inherited from the model
- Input layers: Not specified, inherited from the model
- Output layers: Not specified, inherited from the model
- Input shapes: [1,416,416,3]
- Mean values: Not specified
- Scale values: Not specified
- Scale factor: Not specified
- Precision of IR: FP16
- Enable fusing: True
- Enable grouped convolutions fusing: True
- Move mean values to preprocess section: False
- Reverse input channels: False
TensorFlow specific parameters:
- Input model in text protobuf format: False
- Path to model dump for TensorBoard: None
- List of shared libraries with TensorFlow custom layers implementation: None
- Update the configuration file with input/output node names: None
- Use configuration file used to generate the model with Object Detection API: None
- Operations to offload: None
- Patterns to offload: None
- Use the config file: D:\ML\tensorflow-yolo-v3-master\yolo_v3_tiny_mymodel.json
Model Optimizer version: 2019.3.0-408-gac8584cb7
[ ERROR ] 2 elements of 16 were clipped to infinity while converting a blob for node [['detector/yolo-v3-tiny/Conv/BatchNorm/beta/read/Output_0/Data__const']] to <class 'numpy.float16'>.
For more information please refer to Model Optimizer FAQ (https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html), question #76.
[ ERROR ] List of operations that cannot be converted to Inference Engine IR:
[ ERROR ] FusedBatchNormV3 (11)
[ ERROR ] detector/yolo-v3-tiny/Conv/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_1/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_2/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_3/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_4/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_5/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_6/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_7/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_10/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_11/BatchNorm/FusedBatchNormV3
[ ERROR ] detector/yolo-v3-tiny/Conv_8/BatchNorm/FusedBatchNormV3
[ ERROR ] Part of the nodes was not converted to IR. Stopped.
For more information please refer to Model Optimizer FAQ (https://docs.openvinotoolkit.org/latest/_docs_MO_DG_prepare_model_Model_Optimizer_FAQ.html), question #24.

please tell me how to solve this problem. I have attached my model in this post.

[David_C_Intel]

Hi teeradet,

Thanks for reaching out. Could you please try converting your model using the latest version of OpenVINO™ toolkit (2020.4)?

Let us know if the issue persists.

Regards,

David C

[David_C_Intel]

Hi teeradet,

This thread will no longer be monitored, as we got no response from you. If you have additional questions, please create a new discussion and we will help you there.

Regards,

David C.