use in google colab
The weather image classification problem involving five types of weather cloudy, rainy, sunny, snowy, foggy wasaddressed using CNN with transfer learning. Initial experimentswith Canny Edge Detection and CNN yielded suboptimal resultswith train accuracy at 90.69%, valid accuracy at 72.12%, and test accuracy at 70.34%. Replacing Canny Edge Detection with a CNN using more Conv2D layers improved performance,achieving train accuracy of 95.15%, valid accuracy of 80.64%, and test accuracy of 79.04%. Further enhancement with fine�tuned ResNet yielded the best results, achieving train accuracyof 96.2%, valid accuracy of 91.68%, and test accuracy of 91.8%.