Wall Segmentation

## Overview

Wall Segmentation is a dataset for instance segmentation tasks - it contains Wall NotWall annotations for 508 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

KP SS Indoor Wall Segmentation

## Overview

KP SS Indoor Wall Segmentation is a dataset for instance segmentation tasks - it contains Indoor Walls annotations for 320 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Here are a few use cases for this project:

1. Architecture Analysis: This model could be used by architects or engineers to analyze existing outdoor wall structures in urban or suburban areas, providing insight on patterns, trends, and variations in design for future projects.

2. Real Estate Appraisals: Appraisers could use this model to more accurately estimate property values by rapidly categorizing and analyzing different types of exterior wall materials (brick, wood, vinyl, etc.) and their extent across a property.

3. Urban Planning: Municipalities could harness this model's capabilities to identify and segment the layout of homes or buildings, helping them plan urban developments, design road networks or conduct a census.

4. Augmented Reality Applications: AR developers could use this technology to more accurately overlay digital renderings or data upon real-world physical structures like walls in their applications.

5. Home Renovation: DIY enthusiasts or home improvement professionals could use this model to better plan their projects, by accurately estimating the surface area of their external walls for paint, tiles, or other materials.

Segmentation(wall,door,window)

## Overview

Segmentation(wall,door,window) is a dataset for instance segmentation tasks - it contains Walls Doors Windows annotations for 405 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

Wall_Segmentation

## Overview

Wall_Segmentation is a dataset for instance segmentation tasks - it contains Walls annotations for 740 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [BY-NC-SA 4.0 license](https://creativecommons.org/licenses/BY-NC-SA 4.0).

This is a curated mini-dataset of interior wall images intended for use in generative AI and computer vision experiments, especially within the context of home renovation tools.

The images were carefully selected from open sources (Unsplash and Pexels), and depict clean wall layouts in various room configurations. Each image features visible vertical surfaces and often includes recognizable objects (e.g., tables, beds) to allow contextual size estimation.

This dataset was prepared as part of a capstone project for the Kaggle GenAI Intensive Course 2025Q1, aimed at building a proof-of-concept AI assistant to estimate painting needs based on user-uploaded photos and LLM guidance.

✅ Licensing:

All images in this dataset are sourced from Unsplash and Pexels, both of which allow free use for commercial and non-commercial purposes with no attribution required. This dataset is shared under the CC0 Public Domain license.

📌 Use Cases:

• AI-based renovation estimators
• Wall segmentation experiments
• Context-aware GenAI apps
• Interior design inspiration

Validation dataset and reference code for Carotid Vessel Wall Segmentation and Atherosclerosis Diagnosis Challenge, MICCAI 2022.

Please refer to our website: https://vessel-wall-segmentation-2022.grand-challenge.org/.

Reproducibility of airway dimensions on non-enhanced CT.

Wall Segmentation Model (U-Net++)

Binary segmentation model for detecting walls in floor plans.

  Model Details

Architecture: U-Net++ with EfficientNet-B0 encoder Input Size: 512x512 Dataset: CubiCasa5K Best IoU: 0.7430

  Usage

import torch import segmentation_models_pytorch as smp from huggingface_hub import hf_hub_download

Download model

model_path = hf_hub_download( repo_id="YOUR_USERNAME/wall-segmentation", filename="wall_unet_best.pt" )

Load… See the full description on the dataset page: https://huggingface.co/datasets/JessiP23/5kdataset-v2.

Regional wall motion assessment is critical in the diagnosis of coronary artery diseases, and is usually performed using echocardiography images in clinical practice. However, manual assessment of regional wall motion is time-consuming and requires expertise.

Thus, we published SegRWMA, a segmentation dataset for automatic segment-level assessment of regional wall motion abnormality

Our dataset consists of 198 patients where for each patient three views (A4C, A3C, and A2C) in three modes (2D mode, MCE mode, and LVO mode). Thus, there are totally 1,782 echocardiography videos, with the varying frame size of 640480(71-510). A total of 9881 frames of echocardiography images in three modalities are collected, in which, there are 3,091, 3,391 and 3,399 frames in the 2D mode, LVO mode, and MCE mode, respectively. For each video, six frames (two end-systolic frames, two end-diastolic frames, and two frames between between the end-diastolic and end-systolic frames) are selected for annotation, and the regional wall contour of the left ventricular is labeled as shown in Fig. \ref{fig_segment}. A total of 3,564 segments are labeled in all the 198 patients, among which 45 segments are abnormal. Note that such low incidence rate is a reflection of the real-life statistics of clinical practice at our center. All labels were annotated by four experienced sonographers, and each taking about 5 minutes to finish.

If you used our dataset, please consider to cite our paper in BMVC 2023, "Enhance Regional Wall Segmentation by Style Transfer for Regional Wall Motion Assessment".

HIGHLIGHT 20231101: We have deployed the dataset on Kaggle! https://www.kaggle.com/xiaoweixumedicalai/datasets

Please send emails to xiao.wei.xu@foxmail.com for any questions.

Maximum vessel attenuation in contrast enhancement phases.

Influence of contrast material on combined airway analysis.

Wall Floor2

## Overview

Wall Floor2 is a dataset for instance segmentation tasks - it contains Wall Floor2 annotations for 963 images.

## Getting Started

You can download this dataset for use within your own projects, or fork it into a workspace on Roboflow to create your own model.

  ## License

  This dataset is available under the [CC BY 4.0 license](https://creativecommons.org/licenses/CC BY 4.0).

The GlaS@MICCAI'2015: Gland Segmentation dataset used in the GlaS@MICCAI'2015 challenge consists of 165 images derived from 16 H&E stained histological sections of stage T3 or T4 colorectal adenocarcinoma. The T in TNM cancer staging refers to the spread of the primary tumour). In colorectal cancer, stage T3 means the tumour has grown into the outer lining of the bowel wall, whereas stage T4 means the tumour has grown through the outer lining of the bowel wall. The cancer stage is different from the tumour histologic grade, as the latter indicates the aggressiveness of the tumour. Each section belongs to a different patient, and sections were processed in the laboratory on different occasions. Thus, the dataset exhibits high inter-subject variability in both stain distribution and tissue architecture. The digitization of these histological sections into whole-slide images (WSIs) was accomplished using a Zeiss MIRAX MIDI Slide Scanner with a pixel resolution of 0.465µm. The WSIs were subsequently rescaled to a pixel resolution of 0.620µm (equivalent to 20× objective magnification).

This dataset contains Coronary Computed Tomography Angiography (CCTA) images with annotations for multiple cardiovascular structures.
CCTA is a non-invasive imaging method that provides high-resolution 3D views of the heart and blood vessels.
The dataset was created to support research in automated segmentation of cardiovascular structures, a process that normally requires a lot of manual work from experts.

By including pixel-level labels for different parts of the heart and major vessels, the dataset allows researchers to train and test deep learning models for medical image segmentation.
The goal is to reduce the need for time-consuming manual input and provide consistent, automated results that can help in the diagnosis and monitoring of cardiovascular disease (CVD).

Annotated Structures

Each folder below contains a sequence of images that together represent a 3D volume of the corresponding cardiovascular structure:

• Coronary sinus
• Descending aorta
• Inferior vena cava
• Left atrial appendage
• Left atrial wall
• Papillary muscle – LV
• Posterior mitral leaflet
• Proximal ascending aorta
• Pulmonary artery
• Right ventricular wall
• Superior vena cava

Please see the paper "Digging roots is easier with AI" how the images were captured and the manual counting was performed for the three destructive sampling procedures.

A unitised curtain wall is a type of exterior wall for high-rise buildings, which is comprised of prefabricated modules that hang from the building floor slabs. This dataset depicts a partially installed unitised curtain wall. The dataset consists of

140 images depicting a partially installed unitised curtain wall The camera calibration parameters Measurement of the pose (position and orientation) of the camera with respect to the wall Ground truth images for 60 images from the dataset, segmented as [glass, frame, other]

The dataset is primarily intended to be used in the development of systems to automatically measure the relative pose between the camera and the wall; systems to identify the location where the next curtain wall module should be installed; and related user interfaces.

The Building 4.0 CRC makes no warranty with regard to the accuracy of the information provided and will not be liable if the information is inaccurate, incomplete or out of date nor be liable for any direct or indirect damages arising from its use. The contents of this publication should not be used as a substitute for seeking independent professional advice.

Baseline characteristics by dataset.

The Demand and Sales Analysis of Paper Cup in Western Europe is estimated to be valued at USD 2.4 billion in 2025 and is projected to reach USD 3.3 billion by 2035, registering a compound annual growth rate (CAGR) of 3.3% over the forecast period.

ndimensions/sementic-segmentation-test

Semantic segmentation dataset in COCO format with direct index masks.

  Classes (412 total)

ID Name Color (RGB)

0 background (0, 0, 0)

1 container (57, 107, 229)

2 dried fruit (135, 210, 31)

3 cookie (191, 66, 175)

4 fruit (57, 229, 200)

5 almond (210, 128, 31)

6 bread (97, 66, 191)

7 meatball (64, 229, 57)

8 pineapple (210, 31, 91)

9 broccoli (66, 144, 191)

10 giraffe (215, 229, 57)

11 tree (172, 31… See the full description on the dataset page: https://huggingface.co/datasets/ndimensions/sementic-segmentation-test.