vicariousinc / science_rcn

Reference implementation of a two-level RCN model
https://www.vicarious.com/Common_Sense_Cortex_and_CAPTCHA.html
MIT License
663 stars 195 forks source link

Fail to reproduce results #3

Closed MateusZitelli closed 7 years ago

MateusZitelli commented 7 years ago

The experiments fails to achieve the announced accuracy:

python science_rcn/run.py --train_size 100 --test_size 20 --parallel
INFO:__main__:Training on 10 images...
INFO:__main__:Testing on 20 images...
Total test accuracy = 0.1
tedivm commented 7 years ago

Can you pull down a fresh clone and try this again, showing us the whole terminal session? The command you are running doesn't match the output- the training size is 100 in the command but is only 10 in the output itself (that being said even with train_size=10 we are getting an accuracy of 0.55).

MateusZitelli commented 7 years ago

Cloning and running the example again solved the issue, sorry for the false alarm.