Structure of the repo:
imagenet_table.csv
is the dataset copied manually from the Papers with Code ImageNet Classification section.
generate_frames.py
is a script which takes in the imagenet_table
and generates the following files:
-
top_results_at_time_t.csv
: a file containing a subset ofimagenet_table.csv
, containing only the results that were the state of the art (on 1% accuracy) at the time they were published. I copied its results to the Top Results at Time t sheet in the Drive. -
imagenet_table.csv
: a file containing three columns:- date: all dates from earliest one present in
imagenet_table.csv
to 2019-11-25 - No extra training data: a result from
top_results_at_time_t.csv
published on that date on empty - Extra training data: a result from
top_results_at_time_t.csv
published on that date on empty
I copied its results to the Accuracy to Plot sheet in the Drive.
- date: all dates from earliest one present in
CIFAR-100 Image Classification link
- Double check current results on Papers with Code
- Download sheet "Data with Paper Links" as csv
- Run
generate_frames.py
- Upload/copy
top_results_at_time_t_CIFAR.csv
to CIFAR-100 - Top results at time t sheet on drive - Upload/copy
accuracy_to_plot_CIFAR.csv
to CIFAR-100 - Percentage Correct data sheet on drive --> CIFAR-100 - Top results at time t + details automatically updates