Skip to content

Latest commit

 

History

History
110 lines (81 loc) · 4.91 KB

File metadata and controls

110 lines (81 loc) · 4.91 KB

retinanet-tf

Use Case and High-Level Description

RetinaNet is the dense object detection model with ResNet50 backbone, originally trained on Keras*, then converted to TensorFlow* protobuf format. For details, see paper, repository.

Steps to Reproduce Keras* to TensorFlow* Conversion

  1. Clone and install the original repository (tested on 47fdf189 commit) following the installation steps. Make sure you ran python setup.py build_ext --inplace from the root of repository.
  2. Download the original model from here
  3. Get conversion script:
    1. Get conversion script from repository:
    git clone https://github.com/amir-abdi/keras_to_tensorflow.git
    1. Checkout the commit that the conversion was tested on:
    git checkout c841508a88faa5aa1ffc7a4947c3809ea4ec1228
    
    1. Apply keras_to_tensorflow.patch from <omz_dir>/models/public/retinanet-tf:
    git apply keras_to_tensorflow.patch
    
    1. Run script:
    python keras_to_tensorflow.py --input_model=<model_in>.h5 --output_model=<model_out>.pb
    

NOTE: The conversion was tested with keras==2.3.1, tensorflow==1.15 and h5py<3.0.0.

Specification

Metric Value
Type Object detection
GFlops 238.9469
MParams 64.9706
Source framework TensorFlow*

Accuracy

Metric Value
coco_precision 33.15%

Input

Original Model

Image, name: input_1, shape: 1, 1333, 1333, 3, format: B, H, W, C, where:

  • B - batch size
  • H - image height
  • W - image width
  • C - number of channels

Expected color order: BGR. Mean values: [103.939, 116.779, 123.68]

Converted Model

Image, name: input_1, shape: 1, 1333, 1333, 3, format: B, H, W, C, where:

  • B - batch size
  • H - image height
  • W - image width
  • C - number of channels

Expected color order: BGR.

Output

Original Model

  1. Classifier, name: filtered_detections/map/TensorArrayStack_2/TensorArrayGatherV3. Contains predicted bounding boxes classes in a range [1, 80]. The model was trained on Common Objects in Context (COCO) dataset version with 80 categories of objects, 0 class is for background. Mapping to class names provided in <omz_dir>/data/dataset_classes/coco_80cl_bkgr.txt file
  2. Probability, name: filtered_detections/map/TensorArrayStack_1/TensorArrayGatherV3. Contains probability of detected bounding boxes.
  3. Detection box, name: filtered_detections/map/TensorArrayStack/TensorArrayGatherV3. Contains detection boxes coordinates in a format [y_min, x_min, y_max, x_max], where (x_min, y_min) are coordinates of the top left corner, (x_max, y_max) are coordinates of the right bottom corner. Coordinates are rescaled to input image size.

Converted Model

The array of summary detection information, name - DetectionOutput, shape - 1, 1, 300, 7 in the format 1, 1, N, 7, where N is the number of detected bounding boxes. For each detection, the description has the format: [image_id, label, conf, x_min, y_min, x_max, y_max], where:

  • image_id - ID of the image in the batch
  • label - predicted class ID in range [1, 80], mapping to class names provided in <omz_dir>/data/dataset_classes/coco_80cl_bkgr.txt file
  • conf - confidence for the predicted class
  • (x_min, y_min) - coordinates of the top left bounding box corner (coordinates stored in normalized format, in range [0, 1])
  • (x_max, y_max) - coordinates of the bottom right bounding box corner (coordinates stored in normalized format, in range [0, 1])

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.