Open-Source Network Model Decomposition Data Reference

The accuracy specification is top1 ACC(%) for an image classification network, mAP(%) for an object detection network, or DSC(%) for an image segmentation network. The fine-tune learning rate decreases from 0.1 times of the original learning rate.

Model	Task Type	Datasets	Baseline Precision	Precision After Decomposition	Fine-tune precision after decomposition
ResNet18	Category	ImageNet	70.66	44.02	70.34
ResNet34	Category	ImageNet	74.2	54.92	74.15
ResNet50	Category	ImageNet	75.6	73.64	75.91
ResNet101	Category	ImageNet	78.52	76.97	78.24
InceptionV3	Category	ImageNet	77.98	76.95	77.78
SSD	Detection	coco2017	27.2	24.2	27.9
faster-rcnn	Detection	coco2017	32.5	31	32.2
mask-rcnn	Detection	coco2017	37.9	36.8	38
UNet	Segmentation	SSTEM	87.63	85.05	87.57

Parent topic: See Also