ImRCIFAR is a course* project focused on image classification using the CIFAR-10 dataset. The project aims to explore various traditional machine learning techniques for classifying images across ten distinct categories.
*(CSL2050: PRML, Spring '24, IIT Jodhpur)
The dataset (CIFAR-10) consists of 60000 (32x32) colour images across 10 distinct classes, each class containing 6000 images. The goal is to analyze and apply classical machine learning algorithms to accurately classify these images followed by a retrieval.
- Implement and compare various machine learning techniques, preferably from scratch.
- Evaluate the performances of different algorithms followed by a thorough analysis and choose the best one.
- Providing insights into the strengths and weakness of the approaches tried.
- Performing a failure-case-analysis.
- Implementations: Contains the from-scratch implementations of techniques implemented.
- assets: Stores additional resources such as images, pre-trained model weights, etc.
- requirements.txt: Project dependencies.
- Results: Stores output files, logs, metrics, etc.