[jat10] deepcognition.io soda bottle competition (10h)

[shadiakiki1986]

Hi @jat10

Let's track our progress on the deepcognition.ai competition here

1. [x] warm up (~2h)
   • [x] play around with deepcognition.ai
   • [x] check out the toilet dataset
2. [x] version 0 of our model (~8h)
   • [x] replicate their toilet dataset model on the soda bottles as is and try to get something that is better than random chance (i.e. with 8 labels, accuracy should be > 12.5 %)
     • achieved accuracy of 14% on validation set (built from 90%/5%/5% split train/validation/test)
   • [x] make 1st submission and check score
     • submission opens on July 29
     • submission scores were 0.4 and 0.5
3. [ ] ~version 1 of our model (0h)~
   • [ ] ~add VGG as base layers to benefit from transfer learning. Accuracy here should be better than above 1st model~
   • [ ] ~submit and check score~

[shadiakiki1986]

@jat10 and I just faced an "internal server error" on deepcognition.io. We sent a message to support. Let's see what they say

Edit: they replied. It was a memory issue because we're trying small instances. Reducing to batch size = 1 worked.

[shadiakiki1986]

@jat10 can you please write down here in this issue your progress so far?

[jat10]

@shadiakiki1986 validation accuracy 99.703%

[shadiakiki1986]

That's fantastic! What was your submission score with this model?

On Wed, Jul 4, 2018, 02:10 Jad Tarabay notifications@github.com wrote:

@shadiakiki1986 https://github.com/shadiakiki1986 i don't think someone will get higher than 99.703% Congrats the first prize

[image: screen shot 2018-07-04 at 2 08 11 am] https://user-images.githubusercontent.com/6978243/42248634-42c26624-7f2f-11e8-9bb5-1818fa55d7ca.png

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/teamshadi/ssoc18/issues/2#issuecomment-402317141, or mute the thread https://github.com/notifications/unsubscribe-auth/AIAOhDqNCMx_Pq2HNN9dUm3GNt6stZkhks5uC_nQgaJpZM4U0tn2 .