The primary objectives of this project are: 

• To build a Convolutional Neural Network (CNN) for image classification. 

• To understand how CNNs can be applied to real-world problems. 

• To optimize the CNN model using techniques like data augmentation and regularization to improve its accuracy. 

• To evaluate the model using various metrics such as accuracy, precision, recall, and F1-score.

One of the key metrics for evaluating a machine learning model is its accuracy on both the training and testing datasets. Visualizing the accuracy over epochs is crucial for understanding the model's learning process. A line chart can be plotted to display both training and test accuracy as the number of epochs increases. The X-axis represents the number of epochs (e.g., 1, 2, 3, ..., 50), while the Y-axis shows the accuracy, which is the percentage of correct classifications. In this graph, training accuracy typically increases steadily as the model learns from the training data. On the other hand, the test accuracy may increase initially but could plateau after a certain number of epochs. If the test accuracy diverges from the training accuracy, it may signal overfitting, where the model has learned the training data too well but struggles to generalize to unseen data. If the model's test accuracy closely follows its training accuracy, it indicates that the model generalizes well and performs effectively on new data. 

The graph of Model Loss Over Epochs is an important metric to assess the progress of a model during training. The loss function measures the difference between the model’s predictions and the actual labels, indicating how well the model is performing. In the graph, training loss typically decreases over time, showing that the model is improving and making better predictions as it learns from the training data. Similarly, validation loss also tends to decrease initially, reflecting the model’s ability to generalize to unseen data. However, if the validation loss starts to increase while the training loss continues to decrease, it suggests that the model may be overfitting—meaning it is fitting the training data too well but not generalizing to new, unseen data. 

Both training and validation loss decrease over the first several epochs, which shows improvement in model performance. However, if validation loss starts increasing later on, it would signal a potential issue with overfitting, as the model becomes too specialized in the training data and performs worse on validation data. The graph helps to identify if the model is learning effectively and whether it's generalizing well or overfitting. If both losses continue to decrease and stabilize, it indicates a well-trained model. If there's a divergence between the training and validation loss, adjustments to the model may be necessary.

• An Inductive Proximity Sensor is Used To detect the Metal Waste

• When There Is No Metal the data pin’s (of the sensor) voltage is 9V which is divided into 5.4V and 3.6 V and the 5.4 V enters the Arduino’s 6th pin(For safe operation of Arduino, the 9V is divided)

• When there is metal in front of the sensor, the data voltage becomes 0V and thus metal waste is detected

• The Metal Sensor is mounted in the upper Bin like the picture

• When Dry Waste Falls on the sensor then the output of the sensor (‘AO’ pin which goes to the ‘A0’ pin of the ARDUINO) is high voltage (nearly 5V).

• And when the waste is wet then the output voltage from the sensor will be low (nearly 0V). Thus, Wet sensor senses wet waste.

• The Sensor is attached on the flap, so that the waste falls directly on the Flap

• When Wet Waste is detected, the stepper motor acts accordingly to the flow chart discussed before and eventually, the wet waste goes in the wet wastes’ bin.