Renekton Croc plush detector 🐊
Along with my hand drawn arts.
LET’S BUILD A CROCODILE, PLUSH, RENEKTON DETECTOR
(#416) [Path('reptiles/alligator plush/00000000.jpg'),Path('reptiles/alligator plush/00000001.jpg'),Path('reptiles/alligator plush/00000002.jpg'),Path('reptiles/alligator plush/00000003.jpg'),Path('reptiles/alligator plush/00000004.jpeg'),Path('reptiles/alligator plush/00000005.jpg'),Path('reptiles/alligator plush/00000006.jpg'),Path('reptiles/alligator plush/00000007.jpg'),Path('reptiles/alligator plush/00000008.jpg'),Path('reptiles/alligator plush/00000009.jpg')...](#0) []
| epoch | train_loss | valid_loss | error_rate | time |
|---|---|---|---|---|
| 0 | 1.455003 | 0.249432 | 0.084337 | 00:18 |
| epoch | train_loss | valid_loss | error_rate | time |
|---|---|---|---|---|
| 0 | 0.104521 | 0.049535 | 0.024096 | 00:18 |
| 1 | 0.068319 | 0.012980 | 0.012048 | 00:18 |
| 2 | 0.052283 | 0.011862 | 0.000000 | 00:19 |
| 3 | 0.041685 | 0.010840 | 0.000000 | 00:19 |
Result is very good
Let’s test
'Prediction: alligator plush; Probability: 0.9391'Very good. It is very accurate since my drawing of a plush is very realistic.
'Prediction: renekton; Probability: 0.9834'Easily recognizes my drawing of Renekton as well. I guess I’m an artist
'Prediction: renekton; Probability: 0.9674'Expected the model to predict plush becasue I removed the background but it’s too smart. (In dataset a lot of plush had empty white background contrast to lots of Renekton images having dark backgrounds)
'Prediction: alligator plush; Probability: 0.8644'Indeed I am an alligator plush with my fake beard!