Overview: The HiRISE camera aboard the Mars Reconnaissance Orbiter has captured a massive visual record of the Martian surface, encompassing craters, dunes, and ancient geological features. Hidden within this extensive archive are rare anomalous frames, including sensor artifacts, corrupted crops, and imagery contaminated with non-Martian content. Identifying these rare deviations without prior labeling is a critical challenge that reflects the real-world workflow of planetary-imaging pipelines. Developing a deep generative model to compress and represent dominant imagery allows for the unsupervised isolation and explanation of these structural anomalies.
Submission Guidelines: This is an online submission round. Teams are required to submit a well-structured solution answering the specified case questions. The submission should demonstrate clarity, depth of analysis, logical structure, use of evidence, and relevance to the case. Participants may use any programming language or library (e.g., Python with TensorFlow, PyTorch, scikit-learn). However, it is preferable to use Google Colab or Jupyter Notebook as the coding environment. Compile all the files in a structured manner and upload them to GitHub.
Submissions should include: A well-documented .ipynb notebook with code and outputs . A PDF report containing detailed explanations, plots, tables, and results. All figures should be clearly labelled, and final answers should be highlighted. All of this material has to be put in the GitHub repository. A Markdown file documenting the evolution of the model across different versions (at least 3), following the symptom -> diagnosis -> fix structure.
GitHub link Deadline: 7th September 2026
EOD Rules and Regulations: This is a team event that is open to college students only. Each team can have a maximum of 4 participants. The members of the team can be from different colleges. This is an event comprising two rounds. For the first round, a problem statement will be released. Each team can submit multiple entries, and in that case, only the latest entry will be considered. The top three teams will be awarded certificates of excellence. The teams invited for the offline round will receive a participation certificate. Generating Team ID and Participant IDs on nssc.in : Register separately as participants on nssc.in . Each participant will get a participant ID in the format 26-xxxxxx (26 followed by 6 digits). Please note that this does NOT constitute the team ID. Participant IDs of all concerned participants have to be submitted on Unstop, separated by commas. Go to Data Analytics (https://www.nssc.in/events/dataanalytics) and register as a team (if you haven't already). Once your team's registration is confirmed, your team will be assigned a team ID in the format T-xxxxxxx (T followed by 7 digits). This will be the ID associated with your team for the rest of the event, and this is the ID you need to enter under "Team ID" during registration on Unstop. Please note that entering your team ID on Unstop is mandatory to be considered for further selection process. Otherwise, you won't be eligible to qualify for the next rounds, irrespective of your performance in this round.