Indoor Scenes CVPR 2019
This image dataset contains over 15,000 photos depicting interior scenes, divided into 67 distinct categories. It is designed to train scene recognition models in complex environments.
Approximately 15,620 JPG images divided into 67 categories
Open Database License (ODbL)
Description
The dataset Indoor Scenes CVPR 2019 includes 15,620 images divided into 67 categories of interior scenes, such as libraries, corridors, kitchens or hotel rooms. Each image is annotated by category, making it possible to train contextual recognition models in indoor environments.
What is this dataset for?
- Train image classification models for interior applications
- Testing the robustness of convolutional neural networks (CNN) on complex scenes
- Develop navigation or detection tools for robots or embedded systems
Can it be enriched or improved?
Yes. You can add additional metadata (such as brightness, orientation, or detected objects), or combine this dataset with other sets of home scenes. It is also possible to improve the balance of classes or to increase data to enrich learning.
🔎 In summary
🧠 Recommended for
- Computer vision students
- Robotics projects
- CNN training
🔧 Compatible tools
- PyTorch
- TensorFlow
- Keras
- OpenCV
💡 Tip
Use category breakdown to create a custom benchmark or test few-shot learning algorithms.
Frequently Asked Questions
Can this dataset be used in commercial projects?
Yes, as long as the conditions of the ODbL license are met, including the attribution and sharing of derivatives.
Are there additional annotations like detected objects?
No, images are only classified by scene type. Data can be enriched with additional annotations.
Is it suitable for traditional supervised learning?
Yes, each image is clearly associated with a category, making it easy to train supervised classification models.
