image-classification

Enhancement

A discussion in #614 revealed a good place for improvement - we should ensure that input image is continuous upon start of the augmentation pipeline. This could be implemented by adding image = np.ascontiguousarray(image) to image and mask targets.

A proposed place to add this call - somewhere at the beginning of A.Compose.__call__.

At the moment, from API there are two useful columns about tasks:

However, in tabs (views) there are possible columns tasks:created_at a

My actions before raising this issue

• Read/searched the docs
• Searched past issues

Option to select chunk sizes does not exist in the CLI

Expected Beh

Hi,
I need to download the something-to-something and jester datasets. But the 20bn website "https://20bn.com" are not available for weeks, the error message is "503 Service Temporarily Unavailable".

I have already downloaded the video data of something-to-something v2, and I need the label dataset. For the Jester, I need both video and label data. Can someone share me the

Could FeatureTools be implemented as an automated preprocessor to Autogluon, adding the ability to handle multi-entity problems (i.e. Data split across multiple normalised database tables)? So if you supply Autogluon with a list of Dataframes instead of a single Dataframe it would first invoke FeatureTools:

• take the multiple Dataframes (entities) and try to auto-infer the relationship betwee

🚀 Feature Request

ENUMs for the catalyst-based conventions.
As an example, I will use scope.

Motivation

We have a lot of callbacks that used scope and maybe some other parameter that has only in catalyst-based meaning.
And every time we need to do something like that:

if isinstance(scope, str) and scope in [
            "on_batch_end",
            "on_batch_start",
  

with video_demo.py about 20% speed compared to your 1.0 repo. but thanks much for sharing!