composer.models.ssd.ssd300#

SSD 300 architecture in PyTorch adapted from MLCommons.

Based on MLCommons Reference Implementation here

Classes

SSD300

Build a SSD module to take 300x300 image input, and output 8732 per class bounding boxes.
class composer.models.ssd.ssd300.SSD300(num_classes, model_path=None)[source]#

Bases: torch.nn.modules.module.Module

Build a SSD module to take 300x300 image input, and output 8732 per class bounding boxes.

Parameters

  • num_classes (int, optional) โ€“ The number of classes to detect. Default: 80.

  • model_path (str, optional) โ€“ Path to ResNet34 pretrained model weights. Default: None.