The ssd-resnet-34-1200-onnx
model is a multiscale SSD based on ResNet-34 backbone network intended to perform object detection. The model has been trained from the Common Objects in Context (COCO) image dataset. This model is pre-trained in PyTorch* framework and converted to ONNX* format. For additional information refer to repository.
Metric | Value |
---|---|
Type | Detection |
GFLOPs | 433.411 |
MParams | 20.058 |
Source framework | PyTorch* |
Metric | Value |
---|---|
coco_precision | 20.7198% |
mAP | 39.2752% |
Note that original model expects image in RGB
format, converted model - in BGR
format.
Image, shape - 1, 3, 1200, 1200
, format is B, C, H, W
, where:
B
- batch sizeC
- channelH
- heightW
- width
Channel order is RGB
.
Image, shape - 1, 3, 1200, 1200
, format is B, C, H, W
, where:
B
- batch sizeC
- channelH
- heightW
- width
Channel order is BGR
.
NOTE output format changes after Model Optimizer conversion. To find detailed explanation of changes, go to Model Optimizer development guide
- Classifier, name -
labels
, shape -1, N
, contains predicted classes for each detected bounding box in [1, 81] range. The model was trained on Common Objects in Context (COCO) dataset version with 80 categories of object, 0 class is for background. Mapping to class names provided in<omz_dir>/data/dataset_classes/coco_80cl_bkgr.txt
file - Probability, name -
scores
, shape -1, N
, contains confidence of each detected bounding boxes. - Detection boxes, name -
bboxes
, shape -1, N, 4
, contains detection boxes coordinates in format[y_min, x_min, y_max, x_max]
, where (x_min
,y_min
) are coordinates top left corner, (x_max
,y_max
) are coordinates right bottom corner. Coordinates are rescaled to input image size.
- Classifier, shape -
1, 200
, contains predicted class ID for each detected bounding box in [1, 81] range. The model was trained on Common Objects in Context (COCO) dataset version with 80 categories of object, 0 class is for background. Mapping to class names provided in<omz_dir>/data/dataset_classes/coco_80cl_bkgr.txt
file - Probability, shape -
1, 200
, contains confidence of each detected bounding boxes. - Detection boxes, shape -
1, 200, 4
, contains detection boxes coordinates in format[y_min, x_min, y_max, x_max]
, where (x_min
,y_min
) are coordinates top left corner, (x_max
,y_max
) are coordinates right bottom corner. Coordinates are in normalized format, in range [0, 1].
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>
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-MLPerf.txt
.