1

I am working on an image classification problem. There are 876 images in the training and 600 in the test dataset. It is a multi class classification for plants.

Since this is my first CNN problem, I started working with tensorflow and keras to build my model and then started using transfer learning to improve my performance. The best model which I built so far using Keras is getting me a score of 0.68 ( since the test dataset has no labels, I have to upload my results over and they generate the score using log-loss).

Also, I tried using Fastai and Pytorch which significantly improved my single model score to 0.46.

I have used Resnet50 image dataset for both cases but i cannot experiment using Resnet101 or increasing the batch size since my gpu goes out of memory( also tried using kaggle and google colab).

Is their any way which i can try to further improve the performance? I am basically stuck now since i can't experiment by increasing either batch size or using a better model like Resnet101 etc.

I have uploaded both codes at : https://github.com/akshatshreemali/ImageClassification

Akshat Shreemali
  • 11
  • 1
  • What's is the metric you are reporting like 0.46 and 0.68? What is the accuracy you are currently achieving? – user1825567 Feb 28 '20 at 16:12
  • so the metrics are log loss on the test set. However we don't have labels on the test set so we have to upload our predictions and then they calculate the score based on that – Akshat Shreemali Mar 11 '20 at 15:57

0 Answers0