3 ms·
It's not excellent research. All they did is create a naive NN without any feature recognition and noted that it has a high rate of false positives when present
by kevin_thibedeau 12y ago
It's not excellent research. All they did is create a naive NN without any feature recognition and noted that it has a high rate of false positives when presented with random imagery. Duh.
- woah 12y agoReally? From the article: "They tested with two widely used DNN systems that have been trained on massive image databases." Is this wrong?
- kevin_thibedeau 12y agoGIGO. Just because they used a "Deep" architecture doesn't mean it was suitable for the task.
- jessaustin 12y agoDoes your criticism extend to the whole enterprise of facial recognition, or did this research somehow fail to include the secret sauce that makes it work?