Add Shells and Pebbles Detection

Author: omkar3602

Shells and Pebbles Detection/Dataset/README.md

@@ -0,0 +1 @@
+I have fetched the Shells and Pebbles Dataset from, https://www.kaggle.com/datasets/vencerlanz09/shells-or-pebbles-an-image-classification-dataset, for building and developing this Project. You guys can check out the dataset from the given link.

Shells and Pebbles Detection/Images/images_from_dataset.png

@@ -0,0 +1,82 @@
+# Shells and Pebbles Detection
+
+**GOAL**
+
+To classify the images from the Shells and Pebbles Dataset using a Deep Learning approach.
+
+**DATASET**
+
+https://www.kaggle.com/datasets/vencerlanz09/shells-or-pebbles-an-image-classification-dataset
+
+**DESCRIPTION**
+
+The dataset contains images of 2 different classes, namely Shells and Pebbles. There are 1541 and 2743 images of each type respectively. The task is to classify the images by using a deep learning architecture.
+
+**WHAT I HAD DONE**
+
+First I imported all the required libraries and dataset for this project. Then I imported the dataset and split it into training, validation and testing sets in the ratio 70:20:10. Then I proceeded to build the model. 
+
+I have developed 4 deep learning models to classify the images. First I used a ANN to classify the images. But I didn't get a good accuracy. Secondly I used a CNN to classify the images, but the model wasn't performing well. Lastly, I used a CNN with MaxPooling layers but I didn't get a satisfactory accuracy score. 
+
+Then I proceeded with Transfer Learning. I used the [ResNet 50](https://tfhub.dev/google/imagenet/resnet_v2_50/feature_vector/5) model and trained this model on the images from the dataset. I first got some improvements in the accuracy, after training the model I got an accuracy score of 82.55
+
+**MODELS USED**
+
+The models are:
+
+1. Artificial Neural Network (ANN)
+2. Convolutional Neural Network (CNN)
+3. Convolutional Neural Network (CNN) with MaxPooling layer
+4. Transfer Learning Model (ResNet)
+
+**LIBRARIES NEEDED**
+
+* kaggle
+* tensorflow
+* tensorflow_hub
+* split-folders
+* matplotlib
+
+**VISUALIZATION**
+
+### Images of different Rice classes
+![Images of different Rice classes](../Images/images_from_dataset.png)
+
+### Model 1 (ANN Model) performance graphs
+![Model 1 (ANN Model) performance graphs](../Images/model_1_plot.png)
+
+### Model 2 (Basic CNN Model) performance graphs
+![Model 2 (Basic CNN Model) performance graphs](../Images/model_2_plot.png)
+
+
+### Model 3 (CNN Model with MaxPooling layer) performance graphs
+![Model 3 (CNN Model with MaxPooling layer) performance graphs](../Images/model_3_plot.png)
+
+
+### Model 4 (Transfer Learning Model (ResNet)) performance graphs
+![Model 4 (Transfer Learning Model (ResNet) 20 epochs) performance graphs](../Images/model_4_plot.png)
+
+### Prediction
+
+Prediction on a random image from the testing dataset.
+
+![Prediction](../Images/prediction.png)
+
+**ACCURACIES**
+
+| Model         | Architecture              | Accuracy in % (on testing data) |
+| ------------- |:-------------------------:|:-------------:|
+| Model 1       | ANN Model                 |63.95          |
+| Model 2       | Basic CNN Model           |70.69          |
+| Model 3       | CNN Model with MaxPooling |72.79          |
+| Model 4       | Transfer Learning Model (ResNet) 20 epochs |82.55          |
+
+**CONCLUSION**
+
+I was successfully able to develop a Deep Learning model that can classify images from the given Shells and Pebbles dataset of 2 classes.
+
+**Omkar Jahagirdar**
+
+Connect with me on Linkedin: https://www.linkedin.com/in/omkar-jahagirdar/
+
+Check out my Github profile: https://github.com/omkar3602

Shells and Pebbles Detection/Images/model_1_plot.png

@@ -0,0 +1,5 @@
+kaggle
+tensorflow
+tensorflow_hub
+split-folders
+matplotlib