This model is an instance segmentation network for 80 classes of objects. It is a Mask-RCNN-like model with ResNeXt152 backbone and Feature Pyramid Networks block for feature maps refinement.
| Metric | Value |
|---|---|
| MS COCO val2017 box AP (max short side 480, max long side 640) | 38.9% |
| MS COCO val2017 mask AP (max short side 480, max long side 640) | 34.7% |
| MS COCO val2017 box AP (max height 480, max width 640) | 38.6% |
| MS COCO val2017 mask AP (max height 480, max width 640) | 34.3% |
| Max objects to detect | 100 |
| GFlops | 354.274 |
| MParams | 143.444 |
| Source framework | PyTorch* |
Average Precision (AP) is defined and measured according to standard MS COCO evaluation procedure.
- name:
im_data, shape: [1x3x480x640] - An input image in the format [1xCxHxW]. The expected channel order is BGR. - name:
im_info, shape: [1x3] - Image information: processed image height, processed image width and processed image scale w.r.t. the original image resolution.
- name:
classes, shape: [100, ] - Contiguous integer class ID for every detected object, '0' for background, i.e. no object. - name:
scores: shape: [100, ] - Detection confidence scores in range [0, 1] for every object. - name:
boxes, shape: [100, 4] - Bounding boxes around every detected objects in (top_left_x, top_left_y, bottom_right_x, bottom_right_y) format. - name:
raw_masks, shape: [100, 81, 28, 28] - Segmentation heatmaps for all classes for every output bounding box.
[*] Other names and brands may be claimed as the property of others.