{ "fetched_at": "2026-10-20T03:42:02Z", "source_url": "https://www.kaggle.com/competitions/brouillonpourmescompetitions/data", "pages": [ { "name": "data-description", "content": "## Data:\nIn this competition, participants will have access to a meticulously curated dataset comprising 128 images of vias, each measuring 129x129 pixels in size, **note that in all the provided images the via is always located at the center of the image**. These images have been carefully selected to contain the characteristics of the \"Alignment Offset Error\".\n\n## Ground Truth Data:\nAccompanying the image dataset is a CSV file containing ground truth values for the evaluated area of the Alignment Offset Error, measured in **pixels**. This CSV file is linked with the names of the corresponding images, this provides the participants with the ability to use these labeled data to train a model if desired, to facilitate the process for the participants we have already split the data into training and testing subsets, that can be found in the following files:\n\n* **train.csv** - the training set\n* **test.csv** - the test set\n* **sample_submission.csv** - a sample submission file in the correct format. **note that you are required to provide values for the area of all provided images.**\n\n**Columns**\n\n* `index` - name of the image\n* `area` - calculated area of Alignment Offset Error\n\n**It is important to note that the CSV file named \"Ground_Truth.csv\" and therefore the combination of the \"training.csv\" and \"testing.csv\" contains ground truth values for only 102 images out of the total 128 images in the dataset**. This intentional omission of ground truth values for certain images serves two primary purposes: to facilitate model evaluation and validation and to discourage overfitting by penalizing models that rely excessively on the provided data. \n\n## Data Format:\nParticipants are welcome to utilize the provided CSV file to read the images as input data and extract the corresponding ground truth values for the evaluated area of Alignment Offset Error. The ultimate goal of this competition is to through the development of an algorithmic approach or training a machine learning model to predict the area of the Alignment Offset Error in nanometers within the images accurately.\n\n## Image Format:\nAll images in the dataset are encoded in JPEG format, ensuring compatibility and ease of access for participants across various computational environments and platforms. \n\n## Conversion factor:\n**The Ground Truth data provide the Misalignment error Area in pixels, in order to convert it and therefore get the area in nanometers participants can multiply their predictions by the current conversion factor:**\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F7429498%2Fc3b4203b1ca89f97f916be52103a3786%2FCodeCogsEqn.gif?generation=1709653147371695&alt=media)\n\n## The Expected Output:\nKeep in mind that what is expected from the participants is to **provide either csv file or notebook** that would generate a file named **\"submission.csv\"** that should contain the predicted area in nanometers **for ALL the provided images**. Please refer to the provided \"sample_submission.csv\" file.\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Description", "content": "During the manufacturing and analysis phases, meticulous attention to TSVs is essential, as several defects can potentially arise, impacting device functionality and reliability.\n\nIn our competition, we will delve into the realm of semiconductor error assessment, focusing on one particular defect: \n**Alignment Offset Error:** This error occurs during the exposure of a lacquer/paint in the lithography process. \nAlignment Offset Error refers to the offset that occurs during the exposure from the upper layer to the lower layer. This is typically due to the precision of the alignment performed during lithography. \nThis specific anomaly manifests as a visible misalignment, often appearing as a dark area within the TSV structure, typically situated at the bottom of the via. However, it is imperative to note that the Alignment Offset Error may also manifest in other locations within the via as you will see later on while navigating through the provided data, highlighting the complexity of error detection and characterization in semiconductor devices.\n\nFor instance, in a well-aligned via with no misalignment errors, the area of the defect should be 0.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F7429498%2Fb79feea9cf1e8de0ae570c89efeb90ea%2FGood.jpg?generation=1708506348499843&alt=media)\nConversely, in a via exhibiting misalignment at the bottom, the area of the defect would likely be greater than 0, \n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F7429498%2F32a3a5f037b77a635ac8b2ae66a84dcc%2Falignment%20err.jpeg?generation=1708506366055979&alt=media)", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "Evaluation", "content": "Participants' submissions will be evaluated based on their ability to accurately predict the area of errors in the provided images. The evaluation metric for this competition will be the **mean squared error.**\n", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "rules", "content": "Participants are expected to adhere to the following rules throughout the competition:\n\n1. All submissions must be original work created by the participant or the participant's team.\n2. Participants must not use any external data sources or pre-trained models not provided by the competition organizers.\n3. Team size should not exceed 3 participants", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false }, { "name": "abstract", "content": "Dear Participants welcome to the main round of the ENAS Winter School Hackathon! \n\n# **Challenge Brief:**\nOne of our primary objectives at the Fraunhofer Institute of Electronic Nanosystems is to use AI to automate and streamline the analysis of metrological measurements, that is because manual analysis can be tedious and error prone.\nTo achieve this goal, we employ state-of-the-art image processing and computer vision methodologies. These advanced techniques enable us to transition from manual analysis to semi-automatic and fully-automatic processes, significantly enhancing the speed, accuracy, and reliability of optical metrological measurements analysis. **A slightly challenging analysis task is to assess errors in metrological scans. one of these errors is the \"Alignment Offset Error\", which will be your task here.**\n\n\n# **What is a Through-Silicon Via (TSV)?**\nImagine a microscopic highway running through the heart of a silicon wafer, connecting different layers of electronic circuits. That's precisely what a TSV is—a vertical channel etched through silicon substrates, allowing for the integration of various components and enhancing the performance and efficiency of semiconductor devices. TSVs play a pivotal role in enabling advanced technologies such as 3D integrated circuits, stacked memory architectures, and micro-electro-mechanical systems (MEMS).\n\n\n# **What to Expect:**\n- Data Exploration: Dive into a curated dataset containing semiconductor images paired with the corresponding Area.\n\n- Algorithm Development: Build and fine-tune your algorithms to predict the Area based on the given dataset.\n\n- Evaluation: Your solutions will be assessed based on their ability to align with the ground truth Area values.", "isPublished": false, "hasIsPublished": false, "postTitle": "", "hasPostTitle": false, "mimeType": "", "hasMimeType": false } ] }