Skip to content

Latest commit

 

History

History
130 lines (83 loc) · 4.54 KB

README.md

File metadata and controls

130 lines (83 loc) · 4.54 KB

yolo-v3-onnx

Use Case and High-Level Description

YOLO v3 is a real-time object detection model in ONNX* format from the repository which is converted from Keras* model repository using keras2onnx converter. This model was pre-trained on Common Objects in Context (COCO) dataset with 80 classes.

Specification

Metric Value
Type Detection
GFLOPs 65.998
MParams 61.930
Source framework ONNX*

Accuracy

Accuracy metrics obtained on Common Objects in Context (COCO) validation dataset for converted model.

Metric Value
mAP 48.30%
COCO mAP 47.07%

Input

Original model

  1. Image, name - input_1, shape - 1, 3, 416, 416, format is B, C, H, W, where:

    • B - batch size
    • C - channel
    • H - height
    • W - width

    Channel order is RGB. Scale value - 255.

  2. Information of input image size, name: image_shape, shape: 1, 2, format: B, C, where:

    • B - batch size
    • C - vector of 2 values in format H, W, where H is an image height, W is an image width.

Converted model

  1. Image, name - input_1, shape - 1, 3, 416, 416, format is B, C, H, W, where:

    • B - batch size
    • C - channel
    • H - height
    • W - width

    Channel order is BGR.

  2. Information of input image size, name: image_shape, shape: 1, 2, format: B, C, where:

    • B - batch size
    • C - vector of 2 values in format H, W, where H is an image height, W is an image width.

Output

Original model

  1. Boxes coordinates, name - yolonms_layer_1/ExpandDims_1:0, shape - 1, 10647, 4, format - B, N, 4, where:

    • B - batch size
    • N - number of candidates

  2. Scores of boxes per class, name - yolonms_layer_1/ExpandDims_3:0, shape - 1, 80, 10647, format - B, 80, N, where:

    • B - batch size
    • N - number of candidates

  3. Selected indices from the boxes tensor, name - yolonms_layer_1/concat_2:0, shape - 1, 1600, 3, format - B, N, 3, where:

    • B - batch size
    • N - number of detection boxes

Each index has format [b_idx, cls_idx, box_idx], where:

  • b_idx - batch index
  • cls_idx - class_index
  • box_idx- box_index

The model was trained on Common Objects in Context (COCO) dataset version with 80 categories of object. Mapping to class names provided in <omz_dir>/data/dataset_classes/coco_80cl.txt file.

Converted model

  1. Boxes coordinates, name - yolonms_layer_1/ExpandDims_1:0, shape - 1, 10647, 4, format - B, N, 4, where:

    • B - batch size
    • N - number of candidates

  2. Scores of boxes per class, name - yolonms_layer_1/ExpandDims_3:0, shape - 1, 80, 10647, format - B, 80, N, where:

    • B - batch size
    • N - number of candidates

  3. Selected indices from the boxes tensor, name - yolonms_layer_1/concat_2:0, shape - 1, 1600, 3, format - B, N, 3, where:

    • B - batch size
    • N - number of detection boxes

Each index has format [b_idx, cls_idx, box_idx], where:

  • b_idx - batch index
  • cls_idx - class_index
  • box_idx- box_index

The model was trained on Common Objects in Context (COCO) dataset version with 80 categories of object. Mapping to class names provided in <omz_dir>/data/dataset_classes/coco_80cl.txt file.

Download a Model and Convert it into Inference Engine Format

You can download models and if necessary convert them into Inference Engine format using the Model Downloader and other automation tools as shown in the examples below.

An example of using the Model Downloader:

omz_downloader --name <model_name>

An example of using the Model Converter:

omz_converter --name <model_name>

Legal Information

The original model is distributed under the Apache License, Version 2.0. A copy of the license is provided in <omz_dir>/models/public/licenses/APACHE-2.0.txt.