2 ms·
“For example, top-1 accuracy in the ImageNet-1000 improved by between 0.5% and 2.7%. And Average Precision (AP) accuracy increased by up to 1.8% for bounding bo
by MichaelZuo 6y ago
“For example, top-1 accuracy in the ImageNet-1000 improved by between 0.5% and 2.7%. And Average Precision (AP) accuracy increased by up to 1.8% for bounding box and 2.2% for semantic mask in MS-COCO.“
Just how significant are these results?
- bradneuberg 6y agoTop-1 results are much harder than top-5, so the fact that they are bumping up top-1 by a few percentage points is pretty significant IMHO. Also getting 1 to 2% bump for the segmentation task is nothing to sneeze at.