CONQUERING FASHION MNIST WITH CNNs USING COMPUTER VISION
- 0 Collaborators
The Fashion-MNIST dataset is used as a standard for assessing how well image classification models perform. Classifying fashion items presents a difficult task that is applicable to real-world applications. I have tried to add to the body of knowledge in the field of computer vision by creating ...learn more
Project status: Published/In Market
oneAPI, Artificial Intelligence, Intel® Unnati
Overview / Usage
Fashion-MNIST is a dataset commonly used for image classification tasks in the field of computer vision. It consists of 60,000 training images and 10,000 testing images, each of which is a grayscale 28x28 pixel image belonging to one of ten different fashion categories. And in this project I have made a CNN architecture which will help us in classifying the clothing items and then I have made use of the optimization provided by intel Devcloud platform to optimize my model and then compared the BASELINE INFERENCE LATENCY and ACCURACY with and without the optimization to ensure that the model is highly optimized.
Methodology / Approach
I basically followed the following steps in order to come up with the final convolution neural network model, I will talk in detail about each step ahead in the report that i will upload along with this . 1. Data Preprocessing : Before we feed the data to our CNN architecture we need to process it as per the requirement of our model . 2. Making the model architecture from scratch : We need to make our Model layer by layer from scratch in order to make it capable of learning and predicting . 3. Model training : After our model is ready we need to train it in order to make it efficient enough to successfully predict with high accuracy. 4. Hyperparameter Tuning : This step is done to find a balance between all the parameters , which ensures highest accuracy and best working of the model without overfitting. 5. Model Evaluation : After our model is trained we need to check its metrics like accuracy and loss to find out how good our model is . Sometimes we also look at precision , recall and confusion matrix. 6. Model optimisation : This step is done to make the model’s inference more accurate and faster. 7. Model Deployment : After a model is ready , we need to deploy it to other platforms to use it .
Technologies Used
Intel Devcloud
Intel oneAPI
Jupyter
Tensorflow
Documents and Presentations
Repository
https://github.com/Ashutoshjha0007/intelunnati_AshutoshJha/tree/main