Skip to content

ROADMAP.md

Latest commit

 

History

History
83 lines (58 loc) · 3.89 KB

ROADMAP.md

File metadata and controls

83 lines (58 loc) · 3.89 KB
Raw

Summary

The project that has been assigned to you is the following: Plant recognition

Step 0

  • Framing (Our first meeting): 29.04.2024

Step 1

  • Data mining + DataViz’: 13.05.2024
  1. Your first task will be to define the context and the scope of the project: I expect you to really take the time to understand the project and to learn as much as possible about the concepts it will introduce.
  2. You will then have to take in hand and discover your dataset and make an almost exhaustive analysis of it in order to highlight the structure, the difficulties and possible biases of the dataset.

You can use this template: Data Audit 3. I will also expect at least 5 graphical representations built from your dataset, visual and especially relevant. For each of them I will expect:

  • A precise commentary, which analyzes the figure and provides a “business” opinion.
  • A validation of the observation by data manipulation or a statistical test.

The report is here.

Step 2

  • Pre-processing and feature engineering : 17.05.2024

Following the observations made in the previous step, you will have to clean your dataset and perform the pre-processing steps. If necessary, you will have to transform and/or add features and enrich your dataset. The objective at the end of this step is to have a dataset ready to perform in-depth analysis and/or modeling using Machine (and Deep) Learning.

Rendering 1: data exploration, data visualization and pre-processing report. After rendering 1, if not already done, I will instantiate a GitHub repo for your group, which will have to respect the template provided and be sent to the jury with the final report.

Step 3

  • Modeling: 14.06.2024

Step 1:

  • 24.05.2024
  1. Elaboration of the first modeling iterations with baseline models.
  2. Analyze your first results, is your approach relevant? And your model?

Step 2:

  • 31.05.2024
  1. Use Machine Learning and statistical metrics to better understand the results of your model.
  2. Optimization: Try to improve your model with optimization tools, you can try other more complex models and compare the results.

Step 3:

  • 14.06.2024
  1. Advanced modeling: Use bagging/boosting and Deep Learning techniques.
  2. Interpretation: Use the interpretability tools to better understand the results of your models.
  3. Scientific and business conclusions based on the success or failure of the modeling.

Step 4

  • 14.06.2024

Rendering 2: Modeling report here.

Step 5

  • Final report + codes on GitHub: 21.06.2024

You will submit the final report: made up of the first two renderings, it concretizes your vision and your work on your project by integrating a conclusion and an opening. You also turn in your clean and commented code on GitHub, if time allows you will be able to refine your model and modeling pipeline during this step.

Step 6

  • Streamlit + Defense: 01.07.2024

The document Defense_Methodology summarizes the defense’s organization.

The presentation will be as follows:

  • 20 minutes of presentation
  • 10 minutes of questions from the jury members

You can make either :

  • a Powerpoint presentation + Streamlit demo
  • the entire presentation with your Streamlit application

Your application will have to :

  • be aesthetically pleasing, with several tabs
  • carefully coded, without re-training your model
  • works without bugs

Note

Special attention will be given to the form of the intermediate and final reports, they must contain illustrations, a layout, no spelling mistakes etc... If the reports are not up to standard and delivered on time, your project will not be validated.