📃: Weather Image Recognition Model

[Avdhesh-Varshney]

🔴 Title : Weather Image Recognition Model
🔴 Aim : Brief approach for this project from a basic level upto highest possible accuracy.
🔴 Brief Explanation :

• Link of the dataset
• Perfoming EDA
• Attach any blog for reading related to the topic
• Explain the features properly
• Add necessary comments for the code understanding
• Attach your notebook links
• Applying atleast 3-4 models and compairing them.

NOTE: Follow this readme template

Screenshots 📷

---

✅ To be Mentioned while taking the issue :

• Full name :
• What is your participant role? (Mention the Open Source Program name. Eg. GSOC, GSSOC, SSOC, JWOC, etc.)

---

Happy Contributing 🚀

All the best. Enjoy your open source journey ahead. 😎

[Swish78]

Weather Image Recognition Model

Aim

The aim of this project is to develop a model that can accurately classify weather conditions based on images, starting from a basic approach and refining it to achieve the highest possible accuracy.

Dataset

Weather Dataset from Kaggle

This dataset contains 6862 images of different types of weather, divided into 11 classes: dew, fog/smog, frost, glaze, hail, lightning, rain, rainbow, rime, sandstorm, and snow.

Performing EDA

Steps:

1. Load the dataset and labels.
2. Visualize sample images.
3. Check for missing values.
4. Generate summary statistics such as the number of images per class and image dimensions.

Blog for Reading

Understanding Image Classification

Features Explanation

Input Features:

• Images: The primary input feature is the image itself.
• Image Dimensions: Typically resized to a standard size (e.g., 128x128 pixels).

Preprocessing Steps:

1. Resizing images to a uniform size.
2. Normalizing pixel values to a range of [0, 1].
3. Applying random transformations such as rotation, zoom, and flip for augmentation.

Models Applied

1. Custom Convolutional Neural Network (CNN)

• Implement a simple CNN with a few convolutional layers.

2. Transfer Learning with VGG16

• Use a pretrained VGG16 model with fine-tuning.

3. Transfer Learning with ResNet50

• Use a pretrained ResNet50 model with fine-tuning.

4. Transfer Learning with DenseNet

• Use a pretrained DenseNet model with fine-tuning.

Loss Function

Refer to this paper for a specialized loss function: A Survey on Loss Functions for Deep Learning Classification.

Comparison of Models

Metrics:

• Accuracy: Primary metric for comparison.
• Precision, Recall, F1-Score: Additional metrics for detailed evaluation.

Full Name:

Swayam Patil

Participant Role:

Open Source Program name - VSOC

[Avdhesh-Varshney]

@Swish78 this is to be mentioned in the readme file during pr. Then, i will review it there. In issues you have to mention only your name and open source program only.

[Swish78]

Understood, I'll update the readme file during the pull request. Thanks!

Swayam