This repository contains assignments for the Fundamental of Data Science (FDS) course. Each homework assignment covers the key topics we'll see during our lectures. This README will be updated as new assignments are added.
| 1. | Image Filtering | It's a fundamental process in image processing used to enhance features, suppress noise, or extract information from an image. It serves as a basis for more complex image analysis tasks. |
|---|---|---|
| Number | Subsection | Description |
| 1.0 | Warm-Up | Introduction to convolution and its properties, helping to set the theorical groundwork of filtering techniques. |
| 1.1 | 1D Filters | Basics of one-dimensional filtering, including Gaussian filters for smoothing and Laplacian filters for edge detection. |
| 1.2 | 2D Filters | Explore two-dimensional filtering for direct application to images, using techniques like Gaussian and Laplacian filtering to enhance, smooth, or emphasize specific image features. |
| 2. | Multi-Scale Image Representations | Tecniques used to analyze images at different levels of detail to capture important features more effectively. |
|---|---|---|
| Number | Subsection | Description |
| 2.1 | Prewitt Operator | Introduces the Prewitt operator, a gradient-based method used for edge detection |
| 2.2 | Canny Edge Detector & Template Matching | Apply the Canny edge detection method, a multi-step process for identifying edges while minimizing noise. |
| 2.3 | Harris Corner Detector | It identifies key points (corners) in an image, making it useful for recognizing distinctive features. |
| 2.4 | Gaussian Pyramid & Aliasing | Gaussian pyramids are used to analyze images at various resolutions. This section also discusses aliasing, which can distort images if high-frequency information is not properly handled during downscaling. |
| 2.5 | Multi-Scale Template Matching | Combining template matching with Gaussian pyramids, to identify objects across scales (varying image resolutions). |
| 3. | Object Identification | Techniques to identify and differentiate objects within images using histogram analysis and similarity metrics. |
|---|---|---|
| Number | Subsection | Description |
| 3.1 | 3D Joint Color Histogram | Foundation for color-based object identification. |
| 3.2 | Types of Histograms | To compare images based on intensity, color, or other feature distributions. |
| 3.3 | Histogram Metrics | Calculate similarity between histograms to assess how closely two images match based on their histogram features. |
| 3.4 | Image Retrieval | To identify and return similar images from a dataset based on key features. |
| 3.5 | Report | Summarize the findings, discussing each step and technique applied, and analyzing the performance and accuracy of each approach. |
| 4. | Performance Evaluation | Evaluating the performance of the implemented techniques and metrics to assess their effectiveness in object detection and identification. |
|---|---|---|
| Number | Subsection | Description |
| 4.1 | Evaluation | Analyze standardized evaluation metrics. |
| 4.2 | Nearest Neighbors | It compares a target image to others in a dataset, ranking them by similarity. |
| 4.3 | Retrieval Metrics | Measure retrieval success using metrics like precision, recall, and F1 score, evaluating the effectiveness of image retrieval methods. |
| 4.4 | Analysis & Report | Provide a final report on the evaluation metrics and results, interpreting how well each method performed and identifying areas for improvement. |
Location: HW1_v3.ipynb