{ "fetched_at": "2026-10-19T07:47:45Z", "source_url": "https://www.kaggle.com/competitions/aoml-hackathon-1/data", "pages": [ { "name": "data-description", "content": "Here is the dataset structure:\n\n- **train**: This folder contains subfolders for each Pokémon, with images in JPG format for training your model.\n- **validation**: This folder contains subfolders for each Pokémon, with images in JPG format for validating your model.\n- **test**: This folder contains all the images (JPG format) for which you need to predict the corresponding class.\n- **sample_submission.csv**: This file serves as a reference for your final submission. The columns should be labeled as follows:\n - **Image_Name**: The file name of each image in the test folder.\n - **Class**: The predicted class (Pokémon name) for each image.\n\nEnsure that your final submission follows the same structure and column names as provided in the **sample_submission.csv file**. **This is purely for you to understand the structure of the dataset, but the final submission must have as many records in the test dataset.**", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "**Welcome to the Pokémon Classification Competition for the AOML Course 2024 at PES University!**\n\nThis team-based competition, hosted on Kaggle, challenges you to build a model that accurately classifies Pokémon based on their features. Your task is to develop a model that can correctly predict the name of a Pokémon using various attributes from the dataset, such as type, stats, and other characteristics. The goal is simple: the team with the highest classification accuracy wins.\n\n**Key details:**\n\n- **Team-based**: This is strictly a group effort, encouraging collaboration and effective teamwork.\n- **Classification task**: Your model must identify the correct Pokémon based on the input features. The dataset consists of images of 143 unique pokemons\n- **Evaluation**: The performance will be judged on accuracy, with the highest accuracy on the test data determining the leaderboard rankings.\n- **Grading**: This competition accounts for 5 marks in your final grade for the course.\n\nThe rubrics have been shared for your reference, and you'll be evaluated not only on your model's accuracy but also on:\n\n1. **Model performance** – accuracy.\n2. **Code quality and reproducibility** – ensuring others can follow your work.\n3. **Documentation and presentation** – a clear explanation of your approach.\n4. **Creativity and innovation** – in feature engineering, model selection, or tuning.\n\nAs the competition progresses, you’ll be able to see how your solution stacks up against others on the Kaggle leaderboard. Make sure to keep iterating, refining, and experimenting with your models to maximize performance.\n\nGood luck!", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "**Evaluation and Submission Guidelines**\nSubmissions for this task will be assessed primarily based on their classification accuracy. The key objective is to accurately predict the pokemon for each image in the test set. The metric used for evaluation is the sklearn.metrics.accuracy_score, which measures the proportion of correct predictions to the total number of samples in the test set.\n\n\n**Leaderboards**\nThere are two leaderboards - a public and a private one. Scores for both are evaluated on different subsets of the solution dataset. The public leaderboard will be dynamic throughout the contest and is subject to change based on submissions. The private leaderboard is not open to all participants during the competition, but will display the final scores once the competition is complete.\n\nThe purpose of the private leaderboard is to act as a proxy to real, unseen, test data. As teams are not exposed to this subset of the data, it encourages them to ensure that their model is generalizing well. If the model is found to overfit on the training sample, or conversely, if the generalizes very well, it will be reflected in the final ranks on the leaderboard.\nTherefore, prepare to see some jumps up and down in your Private Leaderboard ranking - which is truly indicative of the model you've built.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "***Pokémon Classification Competition – Rules of the Game 🏆***\nHey everyone! Ready to flex those brain muscles and classify Pokémon like a pro? This competition’s all about accuracy, speed, and showing off those ML chops. But before you go full Ash Ketchum on this dataset, here are the ground rules. Let’s get this data showdown started!\n\n**1. Squad Goals: Team Formation 👯‍♂️**\n*Join solo, squad up later:* First, join the comp with your individual Kaggle accounts. Once you're in, it’s time to squad up with your ML besties.\n*Team Name Game:* Your team name has to be something cool and follow the format: TeamName_SRN1_SRN2_SRN3_SRN4_SRN5. Only the last three digits of your SRNs.\n*Example:* Gryffindor_001_002_003_004_005. Keep it clever, keep it classy.\n**2. Kernel Notebooks – Your ML Pokéball 💻**\n*One Kernel to Rule Them All:* Every team gets only ONE kernel to create and finesse. Make it count!\n*Kernel Naming Style:* Your kernel must follow the format provided in the Deliverables section under the Overview tab.\n*Example:* Gryffindor_001_002_003_004_005.ipynb.\n**Privacy Mode Activated 🔒:** Keep that kernel private! Share it only with the competition hosts (no spoilers here).\n*Sharing Rules:* All the hosts must be given \"To view\" permissions, and trust us, this is not a suggestion, it’s a rule. The host usernames are in the \"Overview\" tab.\n*Document Like a Boss:* Use markdown to document your wild ride through the world of data. Keep it short, sharp, and snappy:\n- Data insights 📊\n- Model architecture 🏗️\n- Hyperparameter adventures 🔧\n- Results 🏁\n\nRemember, brevity is the soul of wit – we just need the highlights, not a PhD thesis.\n\n**3. Submit or Quit! 🚀**\n*Naming is EVERYTHING:* If your team/kernel name doesn’t follow the rules, your submission’s going straight to the bin. No joke – no proper name, no points.\n*Output file = submission.csv*: This ain’t a debate. If it’s named anything else, Kaggle’s gonna ghost your file.\n*Metadata Match*: Double-check that your submission file’s metadata matches the sample (rows, attributes, and data types). Kaggle won’t even give you a shot if it’s invalid.\n*Multiple submissions = more chances*: Maximum of 20 submissions in a day is allowed. The one with the best predictions wins.\n*Deadline = Game Over*: Once the clock hits zero, no more submissions will be accepted. Kaggle will slam the door, and that’s that.\n*Submissions only via Kaggle*: No emails, no attachments, no funny business – everything goes through Kaggle.\n**4. Plagiarism = Instant KO 🚨**\n*Don’t Copy-Paste Your Way to Victory*: We’re watching. Plagiarism checks are real, and if you get caught, every team involved gets wiped out. Not the kind of game over you want.\n*Keep it Secret, Keep it Safe*: No discussing answers, models, or insider insights on the Kaggle Discussions tab. Play it fair, peeps!\n**5. Leaderboard Glory Awaits! 🎯**\n*Accuracy FTW*: This whole comp is about nailing the right classifications. Highest accuracy wins, and the bragging rights are yours.\n*Code like a Pro*: Clean, structured code is key. We want to be able to run it as smooth as a Pikachu’s Thunderbolt.\n*Documentation* = Marks: Short, sweet, and to the point – tell us why your model is pure fire (without the novel).\n*Innovate and Elevate*: Got a slick idea for feature engineering or a wild model architecture? Show us your stuff and rack up those extra points!\n**Leaderboard**: You’ll get to see where you stand in real time! But beware, only your **public leaderboard score** will be visible. **Your final rank depends on your private test set score, so keep those models robust!**\n\nThat’s it! You’ve got the rules, the dataset, and the chance to become a Pokémon data master. Follow the guidelines, go HAM on that model, and may the best team dominate the leaderboard! Let’s see who can out-train the competition.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "This competition is part of the Algorithms and Optimizations in Machine Learning Course 2024, PES University. The competition is strictly a team event. This competition will account for 5 marks in the final grade. The rubrics have been shared with all the students for their reference.\n\nPlease make sure to go through everything given to you in the Overview, Data and Rules Tabs.\nNote the Deliverables required for grading. If your submissions & notebooks do not follow the specifications, marks will be deducted.\n\nFeel free to contact the hosts for any queries, the details for the same can be found in the Rules section.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }