What is your point of view on accuracy vs complexity? I personally appreciate more models that achieve good acuracy with fewer parameters. Is this something that we should strive for, especially for inference and learning on the edge?
[Papers with Code] Image Classification on ImageNet https://paperswithcode.com/sota/image-classification-on-imagenet …pic.twitter.com/SBsPmkrsVL
-
-
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Sub linear. My fear is that we are meta optimizing for imagenet itself. We need a new bigger dataset.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Which of these novel approaches with low scores that were recently published have potential for SOTA with patient refining and tuning?
-
Case Study: Start with small EfficientNet model, then add params, then add adversarial examples, then add "noise". Although, the dog would appreciate "Drop an Octave".
- Još 1 odgovor
Novi razgovor -
-
-
Is there a benchmark that combines accuracy and number of classes (I.e. how fine grained the categories). I’d be more interested in the increase in detail of the classification.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
Tweet je nedostupan.
-
-
So, the mean top-1 accuracy on ImageNet is about 80 % and has stayed there since at least 2017.
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
Is it me or you are just sharing the leaderboard for people to know about PWC? Sorry it feels like the text of your tweet is missing
Hvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
Čini se da učitavanje traje već neko vrijeme.
Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.