1. Why milestones matter¶
Milestones help you check whether your project is developing at a realistic pace. They translate the semester schedule into concrete project checkpoints.
They are not separate assessment rules. Use them as practical signals: if you are behind, reduce scope, clarify the next decision and ask for feedback.
2. Semester milestones¶
The table below follows the provisional SDS320 schedule. The programme may be adjusted during the semester depending on project progress, student needs and technical requirements.
| Week | Date | Course focus | Project focus | What you should have by then | Suggested mini-output |
|---|---|---|---|---|---|
| 1 | 15 Sep | Course introduction | Project framework | You understand the course structure and project expectations. | 3–5 possible project ideas. |
| 2 | 22 Sep | Project planning | Project topic | You have narrowed your broad interest toward one possible topic. | Draft topic, motivation and first research question. |
| 3 | 29 Sep | Data acquisition | Data & methods + concept prep | You have identified candidate datasets, checked basic fit, and drafted your concept (topic, question, motivation). | Data inventory (source, format, coverage, access route, licence) and a concept-presentation draft. |
| 4 | 06 Oct | Data preprocessing | Concept presentations — Group A | You can present your project concept. | One successful data-loading step. |
| 5 | 13 Oct | Training data | Concept presentations — Group B | You know whether labels, masks, samples or validation data are needed, and have incorporated concept-presentation feedback. | Revised concept note and training/validation plan if relevant. |
| 6 | 20 Oct | Object detection | Data analysis | You are moving from planning into implementation. | First prototype on a small example, tile, subset or sample. |
| 7 | 27 Oct | Semantic segmentation | Data analysis | You have tested whether the selected method fits your data and question. | First model or analysis output, even if incomplete. |
| 8 | 03 Nov | Instance segmentation | Processing pipeline | You can describe the workflow from input data to output. | Workflow diagram and cleaned notebook/script order. |
| 9 | 10 Nov | Image translation | Figures & visuals | You have started creating visuals for data, workflow or first results. | One draft figure or map with a short caption. |
| 10 | 17 Nov | Change detection | Storytelling | You can explain the project argument: question, data, method, result and limitation. | Five-sentence project story. |
| 11 | 24 Nov | Regression | Repository + final-presentation prep | Your repository structure, README and environment information are improving, and you have started preparing your final presentation. | Repository check (README, run order, data instructions) and a final-presentation slide outline draft. |
| 12 | 01 Dec | Segment Anything | Final presentations — Group B | You can present near-final implementation, results and limitations. | Revision list based on presentation feedback (what to fix in report, repository, figures). |
| 13 | 08 Dec | Satellite embeddings | Final presentations — Group A | You have presented final progress and know what needs revision. | polishing your report + repository submission, due Mon 14 Dec, 17:00.** |
| 14 | 15 Dec | Finale Grande | Final review | — | — |
3. Key assessment moments¶
Concept presentations¶
The concept presentation focuses on your project plan: motivation, research question, planned data sources, intended methods, expected output and possible bottlenecks. Its purpose is to receive feedback early enough to improve the project before most implementation work is completed.
Final presentations¶
The final presentation focuses on what you implemented, found and learned. It should present your workflow, key results, figures or maps, interpretation, limitations, remaining challenges and changes from the initial concept.
Final project submission¶
The project report and repository contribute 30% to the final grade. The report must be submitted as a PDF via the corresponding MS Teams assignment by Monday, 14 December 2026, at 17:00.
The report must include a link to your public GitHub or GitLab repository.
4. Use milestones flexibly¶
Not all projects progress at the same speed. Some need more time for data access. Others quickly produce outputs but need more work on evaluation or interpretation.
When you fall behind, do not simply add more work. First simplify:
smaller study area,
shorter time period,
fewer classes,
fewer input datasets,
simpler baseline method,
one clear output instead of several outputs.
5. Flags & checks¶
| Red flag | First check |
|---|---|
| No clear question after Week 2 | Narrow the topic using Research question. |
| No candidate dataset by concept presentation | Prioritise data access before method complexity. |
| Data exist but cannot be opened | Test a smaller file and check format, path and environment. |
| No prototype by the data analysis phase | Reduce area, files or method complexity. |
| Figures are only planned for the end | Create rough figures early, even if they are not final. |
| Repository is not updated | Start the README and document the current run order. |
| Project story is unclear after first outputs | Write the five-sentence project story and check whether the output answers the question. |
6. Mini task¶
Compare your current project status with the table.
Current week:
What should I have by now?
What do I already have?
What is missing?
What is the next small action?
What can I simplify if needed?7. Key takeaways¶
Milestones help you detect delays early.
Concept presentations are for useful early feedback.
Final presentations should show implementation, not only plans.
The final submission requires a PDF report and public repository link.
Continuous progress is safer than late project assembly.