You can add location information to your Tweets, such as your city or precise location, from the web and via third-party applications. You always have the option to delete your Tweet location history. Learn more
Any better metrics that you might not have come up with along the way? Or sets of metrics?
3/ I should mention @wightmanr has a set of metrics he runs for his models that are much more informative about a model's generalization on classification tasks. I'm just not 100% sure yet if I can use this to predict image to image performance:https://github.com/rwightman/pytorch-image-models/tree/master/results …
This is really interesting question about metrics that capture good downstream performance. I guess one good proxy is to look at tasks that use image model backbones (eg object detection), but would be interested to hear more ideas!
(Same as your experience: with exception of EfficientDet, most top object detection models on test-dev seem to stick to using resnet/resnext rather than other architectures)
Thanks! I'd be quite curious to see how things evolve for you. Metrics are really important, but also quite tricky to determine as tasks diversify
Twitter may be over capacity or experiencing a momentary hiccup. Try again or visit Twitter Status for more information.
