Creating a Dataset from Google Image Search Results

If you wish to create an image classifier and want to use the data from Google Image Search results, and want to exclude some of the images, you can use this bookmarklet gi2ds (drag it to your bookmarks bar and click on it after your search). Then you can click on the images you want to exclude. A list is generated for you with all the relevant image-urls for you to process further.

gi2ds is intened to help you when creating an image dataset based on a google images query. It allows you to exclude images that are not relevant by toggling them on and off by clicking on them. Default is that all images are included. The urls are found in a popup down to the right. To get all available images you need to scroll all the way down for more images to load, also pressing the show more results button and continuing scrolling in order to get all the pictures available.

For more info, the code, see GitHub

Inspiration comes from this years course (v3) where i am attending as an International Fellow. The course will be available to the public in January 2019


Artificial Intelligence Competition Leaderboard

I have not seen this cool leaderboard for AI challenges before.

There are a few very interesting similar competition leaderboards for machine learning such as Kaggle and Numerai. Allenai host right now 4 interesting NLP challenges.

Here is the description of one of the challenges:

OpenBookQA: Open Book Question Answering

OpenBookQA is a new kind of question-answering dataset modeled after open book exams for assessing human understanding of a subject. It consists of 5,957 multiple-choice elementary-level science questions (4,957 train, 500 dev, 500 test), which probe the understanding of a small “book” of 1,326 core science facts and the application of these facts to novel situations. For training, the dataset includes a mapping from each question to the core science fact it was designed to probe. Answering OpenBookQA questions requires additional broad common knowledge, not contained in the book. The questions, by design, are answered incorrectly by both a retrieval-based algorithm and a word co-occurrence algorithm. Strong neural baselines achieve around 50% on OpenBookQA, leaving a large gap to the 92% accuracy of crowd-workers.