Waste Classification System using Convolutional Neural Networks.pptx
- 1. Waste Classification System using Convolutional Neural Networks
- 2. Under the Guidance Of Dr. Divya Kumar
- 3. Group Members âś Abhinav Dixit (20151012) âś Ishaan Rajput (20154086) âś Abhishek Sharma (20154077) âś Harshita Rastogi (20154041) âś John Prasad (20154010)
- 4. Objective âś To develop a system to effectively segregate the collected waste on the basis of different categories using the concepts of artificial neural networks and image processing. âś To create Convolution Neural Network(CNN) model which is a powerful and a deep layered network that helps in the classiďŹcation process along with tuning various parameters to get the highest accuracy. âś To help in eliminating the need of a middle-man for the treatment of waste.
- 5. Motivation âś By designing an autonomous system we can reduce the workforce required for waste management and also perform the task efficiently. âś We can classify the waste into one of the three categories â recyclable, compost and landfill(Non-Recyclable waste). âś Hazardous waste materials which cause harm to humans can easily be disposed off.
- 6. Proposed Work âś To carry out the image classiďŹcation of waste materials using Convolutional Neural Networks(CNN). Given an image of a random waste item, the system seeks to categorize the waste item into one of the 5 mentioned categories. âś Categories for classiďŹcation: âś 1. Metal âś 2. Organic waste âś 3. Container âś 4. Paper âś 5. Plastic
- 7. Artificial Neural Networks âś An ANN is a computational model based on the structure and functions of biological neural networks. âś Information that flows through the network affects the structure of the ANN because a neural network changes - or learns, in a sense - based on that input and output. âś There are weighted connections (correspond to synapses) between simulated neurons where signals it receives (numbers) are summed. âś A signal is sent (fired) if a certain threshold is reached.
- 8. Artificial Neural Networks âś An ANN is typically defined by three types of parameters: âś The interconnection pattern between different layers of neurons âś The learning process for updating the weights of the interconnections. âś The activation function that converts a neurons weighted input to its output activation.
- 9. Artificial Neural Networks âś The model learns by repeating the following steps - âś Feed Forward Algorithm - The input layer of neural network is fed with samples. These sample values will be multiplied with the corresponding weights and added up. The bias will be added to the resulting sum and this sum will be passed to the activation function. âś After the Feedforward the output of the neural network is compared with the target output. The difference between expected output of a neuron and actual output of the same neuron gives the error of that neuron. âś Back Propagation Algorithm - Backpropagation algorithm is applied after the feedforward algorithm in order to propagate the errors in other direction of feed forward and adjust the weights to overcome that error.
- 10. Convolutional Neural Network (CNN) âś Why CNN? âś Image recognition is not an easy task to achieve. In Theory, we can use conventional neural networks for analyzing images, but in practice, it will be highly expensive from a computational perspective. âś For instance an image of more respectable size, e.g. 200x200x3, would lead to neurons that have 200*200*3 = 120,000 weights. âś Convolutional Neural Networks take advantage of the fact that the input consists of images and they constrain the architecture in a more sensible way. The layers of a ConvNet have neurons arranged in 3 dimensions: width, height, depth. âś The neurons in a layer will only be connected to a small region of the layer before it, instead of all of the neurons in a fully-connected manner.
- 11. CNN - Structure âś Any CNN comprises of following layers â âś Convolution Layer (With ReLU) âś Pooling Layer âś Flattening âś Followed by fully connected ANN
- 12. Convolution Layer âś The CONV layerâs parameters consist of a set of learnable filters. âś During the forward pass, we convolve each filter across the width and height of the input volume and compute dot products between the entries of the filter and the input at any position. âś As we slide the filter over the width and height of the input volume we will produce a 2-dimensional activation map that gives the responses of that filter at every spatial position. âś Intuitively, the network will learn filters that activate when they see some type of visual feature such as an edge of some orientation, color patterns, etc. on higher layers of the network.
- 13. ReLU Layer âś The Convolution Layer is followed by ReLU (Rectified Linear Unit) activation function, defined as f(x) = Max(0, x). âś It is also known as ramp function. âś It is used to remove linearity in the output of the CONV Layer.
- 14. Pooling âś The Pooling Layer operates independently on every depth slice of the input and resizes it spatially, using the MAX operation. âś Its function is to progressively reduce the spatial size of the representation to reduce the amount of parameters and computation in the network, and hence to also control overfitting. âś The most common form is a pooling layer with filters of size 2x2 applied with a stride of 2 down-samples every depth slice in the input by 2 along both width and height, discarding 75% of the activations. Every MAX operation would in this case be taking a max over 4 numbers.
- 15. Flattening âś We need to convert the output of the convolutional part of the CNN into a 1D feature vector, to be used by the ANN part of it. This is achieved by Flattening. âś It gets the output of the convolutional layers, flattens all its structure to create a single long feature vector to be used by the dense layer for the final classification.
- 16. Fully Connected ANN âś Neurons in a fully connected layer have full connections to all activations in the previous layer, as seen in Conventional Neural Networks. âś This is proceeded by further adding dense hidden layers and finally computing the values of the Output Layer.
- 17. Dataset used âś The dataset was manually collected for every category from various search engines like Bing, Google and Yahoo through a software called Extreme Picture Finder. âś The dataset contains images that include the waste items in diďŹerent illumination, background colors and various angles. âś The dataset consists of a total of 3,302 images belonging to one of the 5 classes. âś Care was taken to include images of objects as they would be when disposed oďŹ like crushed bottles, crumpled paper, and so on to make the training more robust to deal with real-life images.
- 18. Software tools used âś Python 3 âś Tensor-ďŹow âś Keras âś Cuda (NVIDIA Computing Toolkit) âś Matplotlib âś SciPy
- 19. Description Of Model âś After doing parameter hyper-tuning, we got the following setting to produce the best outcome. âś Pictures were rescaled to a size of 256 X 256 pixels. âś The train-test split ratio was taken to be around 12:1. âś Batch size was chosen to be 16. âś We used Categorical Cross-Entropy as loss function and used Adam Optimizer to reduce it over 30 epochs (repetitions).
- 20. Description Of Model - Layers âś Layer 1 - CNN (followed by Batch Normalization and Max Pooling) âś Filters - 96, Kernel Size - 11 x 11, Strides - 4 âś Activation Function - ReLU âś Layer 2 - CNN (followed by Batch Normalization and Max Pooling) âś Filters â 256, Kernel Size - 5 x 5 âś Activation Function - ReLU âś Layer 3-4 - CNN âś Filters â 384, Kernel Size - 3 x 3 âś Activation Function - ReLU âś Layer 5 - CNN (followed by Max Pooling) âś Filters â 256, Kernel Size - 3 x 3 âś Activation Function - ReLU âś Layer 6 - Flattening
- 21. Description Of Model - Layers âś Layer 7 - Fully Connected Hidden Layer âś Neurons - 2048 âś Activation Function - ReLU âś Layer 8 - Fully Connected Hidden Layer âś Neurons - 2048 âś Activation Function â ReLU âś Layer 9 - Output Layer âś Neurons - 5 âś Activation Function - Softmax
- 22. Model Diagram Inputs 3@256x256 Feature Maps 96@64x64 Feature Maps 256@64x64 Feature Maps 384@64x64 Feature Maps 256@64x64 Feature Maps 384@64x64 Hidden Layer 2048 Hidden Layer 2048 Output Layer 5 Convolution 11x11 Kernel, Max Pooling 2x2 ReLU Convolution 5x5 Kernel, Max Pooling 2x2 ReLU Convolution 3x3 Kernel ReLU Convolution 3x3 Kernel ReLU Convolution 3x3 Kernel, Max Pooling 2x2 ReLU Flatten Fully Connected Fully Connected
- 23. Result Analysis and Conclusion Image Size No of Classes Accuracy Validation Accuracy Loss Validation Loss 256x256 5 0.9848 0.7456 0.0864 2.1231
- 24. Result Analysis and Conclusion The model correctly classiďŹed 11/14 images. It could not classify images 2,4 and 12 correctly.
- 25. Demo
- 26. Future Works âś Localization â In addition to ClassiďŹcation of Waste, we can use techniques such as Bounding Box Localization, Instance Segmentation etc. to locate the classiďŹed object in the image. âś Generalization â Rather than constricting our search among 5 classes we can use pre-trained models (like VGG-16) to generalize this approach over all kinds of waste. The model can be ďŹne-tuned to ďŹt our requirements and provide wider and better results. âś Automation â Build a more sophisticated mechanical system to automatically separate and segregate the cluster of wastes. âś Improving the accuracy rates of the system by adding more training and test samples to the existing dataset.
- 27. Thank you!!