Update README

[[_TOC_]]

## Description

## Description & motivation

This folder contains the work for our final project of the EML Proseminar. The project is about classifying different types of fruit based on their images (:arrow_right: **multi-class image classification**)

:arrow_right: Given any picture, we want to make the correct prediction: *Is it a banana, an apple, a strawberry, ...?* :banana: :apple: :strawberry:

## Structure

## Project Proposal

## Related Work

:arrow_right_hook: [`project_proposal.md`](project_proposal.md)

:construction: We will add the most important information from the project proposal to this README file.

## Data

Your data is ready! 

## Baseline Model Evaluation Results

## Metrics

## Baseline

### Overview

We have implemented two types of baseline models: Random and Majority. These are implemented both as custom models and using scikit-learn's `DummyClassifier`. Our dataset involves classifying one out of 30 classes, with a balanced dataset of about 26,500 data points.

- Performance Lower Than Majority Baseline: This scenario is more alarming because the majority baseline is a very naive model. If a model performs worse than this, it might indicate that the model is doing worse than a naive guess of the most frequent class. This could be due to incorrect model architecture, data preprocessing errors, or other significant issues in the training pipeline.

## Experiments

| Classifier | Best Features | Best Parameters | Best Accuracy | Training Time | 

| ---------- | --------------|-----------------| ------------- | ------------- |

| Random Forest |

| Decision Tree |

| (Gaussian) Naive Bayes | 

#### Feature Engineering

## Classifiers

### Naive Bayes

### Decision Tree

### Random Forest

### CNN

## Experiments and results

### Overview

### Feature Engineering

- we are experimenting with different feature combinations, which can be used as an input parameter for our `classify_images.py` script

| Feature / Filter | Length | Description |

| `hsv` | the images are converted to HSV color space and the HSV values are used as features |

| `sobel` | 7500 | the images are 

### Naive Bayes

```json

param_grid = { 'var_smoothing': np.logspace(0,-20, num=20)}

```

**Poor results** for all experiments with a Naive Bayes classifier :thumbsdown:. The best results are achieved using the HSV + Sobel filters. The accuracy (0.178) though is still better than our baseline.

| Resized | Features | Accuracy (Dev) | Best Parameters | Comments |

| ------- | -------- | -------- | --------------- | ---- |

| 50x50   | No filters | 0.113 | `{'var_smoothing': 5.5 * 10^-6}` | :arrow_right_hook:  [GridSearch results](figures/naive_bayes/grid_search_results_50x50_standard_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_50x50_standard_confusion_matrix_var_smoothing_5.455594781168525e-06.png) |

| 50x50   | HSV + Sobel | **0.178** :thumbsup: | `{'var_smoothing': 4.2 * 10^-8}` |  GridSearch results and confusion matrix, see below |

| 50x50   | HSV only | 0.162 | `{'var_smoothing': 7.0 * 10^-4}` | :arrow_right_hook: [GridSearch results](figures/naive_bayes/grid_search_results_50x50_hsv_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_50x50_hsv_confusion_matrix_var_smoothing_0.0006951927961775605.png)  |

| 100x100 | No filters | 0.113 | `{'var_smoothing': 5.5 * 10^-6}` | :arrow_right: no improvement with more features; :arrow_right_hook: [GridSearch results](figures/naive_bayes/grid_search_results_100x100_standard_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_100x100_standard_confusion_matrix_var_smoothing_5.455594781168525e-06.png)  |

**Further findings:**:

- accuracy on the training set is also never higher than 0.20 :arrow_right: the classifier is not overfitting but also not learning anything

![GridSearch](figures/naive_bayes/grid_search_results_50x50_hsv_sobel_var_smoothing.png)

![Confusion Matrix](figures/naive_bayes/GaussianNB_50x50_hsv_sobel_confusion_matrix_var_smoothing_4.281332398719396e-08.png)

- for some classes, the diagonal is quite bright (e.g. apricots and passion fruits) :arrow_right: the classifier is quite good at predicting these classes

- but we also see that the classifier has a **strong bias** towards some classes (e.g. apricots, jostaberries and passion fruits and figs)

### Decision Tree

### Random Forest

**Feature Combinations:**

![GridSearch](figures/random_forest/grid_search_results_50x50_hsv_sobel_max_depth_10_80.png)

![GridSearch](figures/random_forest/grid_search_results_50x50_hsv-only_max_depth_10_80.png)

### Naive Bayes

```json

param_grid = { 'var_smoothing': np.logspace(0,-20, num=20)}

```

**Poor results** for all experiments with a Naive Bayes classifier :thumbsdown:. The best results are achieved using the HSV + Sobel filters. The accuracy (0.178) though is still better than our baseline.

| Resized | Features | Accuracy (Dev) | Best Parameters | Comments |

| ------- | -------- | -------- | --------------- | ---- |

| 50x50   | No filters | 0.113 | `{'var_smoothing': 5.5 * 10^-6}` | :arrow_right_hook:  [GridSearch results](figures/naive_bayes/grid_search_results_50x50_standard_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_50x50_standard_confusion_matrix_var_smoothing_5.455594781168525e-06.png) |

| 50x50   | HSV + Sobel | **0.178** :thumbsup: | `{'var_smoothing': 4.2 * 10^-8}` |  GridSearch results and confusion matrix, see below |

| 50x50   | HSV only | 0.162 | `{'var_smoothing': 7.0 * 10^-4}` | :arrow_right_hook: [GridSearch results](figures/naive_bayes/grid_search_results_50x50_hsv_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_50x50_hsv_confusion_matrix_var_smoothing_0.0006951927961775605.png)  |

| 100x100 | No filters | 0.113 | `{'var_smoothing': 5.5 * 10^-6}` | :arrow_right: no improvement with more features; :arrow_right_hook: [GridSearch results](figures/naive_bayes/grid_search_results_100x100_standard_var_smoothing.png) and [confusion matrix](figures/naive_bayes/GaussianNB_100x100_standard_confusion_matrix_var_smoothing_5.455594781168525e-06.png)  |

**Further findings:**:

- accuracy on the training set is also never higher than 0.20 :arrow_right: the classifier is not overfitting but also not learning anything

![GridSearch](figures/naive_bayes/grid_search_results_50x50_hsv_sobel_var_smoothing.png)

![Confusion Matrix](figures/naive_bayes/GaussianNB_50x50_hsv_sobel_confusion_matrix_var_smoothing_4.281332398719396e-08.png)

- for some classes, the diagonal is quite bright (e.g. apricots and passion fruits) :arrow_right: the classifier is quite good at predicting these classes

- but we also see that the classifier has a **strong bias** towards some classes (e.g. apricots, jostaberries and passion fruits and figs)

## :computer: Usage of the `classify_images.py` script

### CNN

### Set-Up

### Final results

Create a virtual environment and install the required packages:

Having tested different feature combinations, hyperparameters and picture sizes on the development set, we have found our optimal parameters for the final tests on the **test set**.

```bash

python3 -m venv venv

source venv/bin/activate

pip install -r requirements.txt

```

### Run the script

### Feature Importance

```bash

python classify_images.py --classifier <classifier_name> --filters <filters> --resize <resize-1> <resize-2> [--optimize] [--debug]

```

### Data Reduction

## Challenges & Solutions

## Conclusion

## Contact