The rise of multi-million-item dataset initiatives has enabled data-hungry machine learning algorithms to reach near-human semantic classification performance at tasks such as visual object and scene recognition. Here we describe the Places Database, a repository of 10 million scene photographs, labeled with scene semantic categories, comprising a large and diverse list of the types of environments encountered in the world. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide scene classification CNNs (Places-CNNs) as baselines, that significantly outperform the previous approaches. Visualization of the CNNs trained on Places shows that object detectors emerge as an intermediate representation of scene classification. With its high-coverage and high-diversity of exemplars, the Places Database along with the Places-CNNs offer a novel resource to guide future progress on scene recognition problems.
Download full-text PDF |
Source |
|---|---|
| http://dx.doi.org/10.1109/TPAMI.2017.2723009 | DOI Listing |
Publication Analysis
Top Keywords
Similar Publications
Transcutaneous Auricular Vagus Nerve Stimulation Enhances Emotional Processing and Long-Term Recognition Memory: Electrophysiological Evidence Across Two Studies.
Psychophysiology
March 2025
Department of Biological Psychology and Affective Science, Faculty of Human Sciences, University of Potsdam, Potsdam, Germany.
Recently, we found that continuous transcutaneous auricular vagus nerve stimulation (taVNS) facilitates the encoding and later recollection of emotionally relevant information, as indicated by differences in the late positive potential (LPP), memory performance, and late ERP Old/New effect. Here, we aimed to conceptually replicate and extend these findings by investigating the effects of different time-dependent taVNS stimulation protocols. In Study 1, an identical paradigm to our previous study was employed with interval stimulation (30-s on/off).
View Article and Find Full Text PDFIMRMB-Net: A lightweight student behavior recognition model for complex classroom scenarios.
PLoS One
March 2025
Department of Computer Science and Information Engineering, Harbin Normal University, Harbin, China.
With the continuous advancement of education informatization, classroom behavior analysis has become an important tool to improve teaching quality and student learning outcomes. However, student classroom behavior recognition methods still face challenges such as occlusion, small objects, and environmental interference, resulting in low recognition accuracy and lightweight performance. To address the above problems, this study proposes a lightweight student behavior recognition model based on Inverted Residual Mobile Block (IMRMB-Net).
View Article and Find Full Text PDFOptimizing multimodal scene recognition through relevant feature selection approach for scene classification.
MethodsX
June 2025
JNN College of Engineering, Shimoga, Karnataka, India.
Scene classification plays a vital role in various computer vision applications, but building deep learning models from scratch is a very time-intensive process. Transfer learning is an excellent classification method using the predefined model. In our proposed work, we introduce a novel method of multimodal feature extraction and a feature selection technique to improve the efficiency of transfer learning in scene classification.
View Article and Find Full Text PDFGaze dynamics during natural scene memorization and recognition.
Cognition
March 2025
Department of Radiology, Washington University in St. Louis, St. Louis, MO 63110, USA. Electronic address:
Humans can rapidly memorize numerous images, which is surprising considering the limited visual sampling of each image. To enhance the probability of recognition, it is crucial to focus on previously sampled locations most likely to support memory. How does the visuomotor system achieve this? To study this, we analyzed the eye movements of a group of neurotypical observers while they performed a natural scene memorization task.
View Article and Find Full Text PDFAutomatic trailer generation for movies using convolutional neural network.
Sci Rep
March 2025
Faculty of Humanities and Arts, Macao University of Science and Technology, Taipa, 999078, Macao, China.
In this paper, a new method for producing movie trailers is presented. In the proposed method, the problem is divided into two sub-problems: "genre identification" and "genre-based trailer production". To solve the first sub-problem, the poster image and subtitle text processing strategy has been used in which, a convolutional neural network (CNN) model has been used to extract features related to the movie genre from its poster image.
View Article and Find Full Text PDFWant AI Summaries of new PubMed Abstracts delivered to your In-box?
Enter search terms and have AI summaries delivered each week - change queries or unsubscribe any time!