Awesome work. Any recommendations for best model in terms of accuracy/flops or accuracy/parameters? It's interesting to see resnet50 significantly higher than some of the deeper resnets and also lots of the simple variations of it.
-
-
-
Thanks. ResNet50 (sometimes w/ 'D' additions or NeXT/SE) is still my goto for new experiments due to a good balance of GPU throughput/memory utilization vs accuracy and ease of training. As you can see, lots of room left for improvements :)
- Još 2 druga odgovora
Novi razgovor -
-
-
When I was waiting an eternity for these validation runs to finish, I had a thought... it'd be great if
@sotabench could run these test sets, with pretty graphs and some interactivity for exploring ranking/accuracy changes. What do you think@rosstaylor90 ?Prikaži ovu nitHvala. Twitter će to iskoristiti za poboljšanje vaše vremenske crte. PoništiPoništi
-
-
-
I see some prefixes like "tf" and "gluon". Your description says pretrained "PyTorch" models. What are they?
-
Some of the models were originally trained in different frameworks and ported to PyTorch. When I did the port, I added a prefix to differentiate.
Kraj razgovora
Novi razgovor -
-
-
Thanks for putting this together! This is a good first step towards properly validating CV models for generalization capabilities. I'm personally a big fan of validating CV models with out of distribution samples.
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.