Photo Recall: Using the Internet to Label Your Photos

Kumar, Neeraj; Seitz, Steve

doi:10.1109/cvprw.2014.120

Cited by 10 publications

(9 citation statements)

References 23 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The model choice with the BIC criterion according to (4) gives the best precision for this dataset. It is achieved in the deep domain constructed by a S2M network with three mixture components (see Table 1).…”

Section: Resultsmentioning

confidence: 99%

“…Such capability is highly beneficial for intelligent systems, especially in certain scenarios, such as open-vocabulary image retrieval that allows users to formulate queries to image collections using arbitrary natural language queries. The use of uncurated image sets obtained from the web, however, is known to be challenging [3,4], since despite the ever-improving performance of image search engines, the returned image sets still contain irrelevant images, since many natural language queries are inherently polysemous, and since many visual concepts often correspond to different visual aspects (e.g. outdoor and indoor views of a certain landmark building).…”

Section: Set2model Networkmentioning

confidence: 99%

“…The work [4] describes a system for indexing user image collections with the help of Internet search and various datasources such as maps to find locations. In [21], a robotic system is proposed that uses various Internet datasources including image search engine to learn how to perform certain complex activities.…”

Section: Related Workmentioning

confidence: 99%

See 2 more Smart Citations

Set2Model Networks: Learning Discriminatively To Learn Generative Models

Kuzmin

Vakhitov

Lempitsky

2017

2017 IEEE International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

We present a new "learning-to-learn"-type approach that enables rapid learning of concepts from small-to-medium sized training sets and is primarily designed for webinitialized image retrieval. At the core of our approach is a deep architecture (a Set2Model network) that maps sets of examples to simple generative probabilistic models such as Gaussians or mixtures of Gaussians in the space of high-dimensional descriptors. The parameters of the embedding into the descriptor space are trained in the end-to-end fashion in the meta-learning stage using a set of training learning problems. The main technical novelty of our approach is the derivation of the backprop process through the mixture model fitting, which makes the likelihood of the resulting models differentiable with respect to the positions of the input descriptors.While the meta-learning process for a Set2Model network is discriminative, a trained Set2Model network performs generative learning of generative models in the descriptor space, which facilitates learning in the cases when no negative examples are available, and whenever the concept being learned is polysemous or represented by noisy training sets. Among other experiments, we demonstrate that these properties allow Set2Model networks to pick visual concepts from the raw outputs of Internet image search engines better than a set of strong baselines.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Set2model Networkmentioning

confidence: 99%

See 1 more Smart Citation

Set2Model Networks: Learning Discriminatively To Learn Generative Models

Kuzmin

Vakhitov

Lempitsky

2017

2017 IEEE International Conference on Computer Vision Workshops (ICCVW)

View full text Add to dashboard Cite

show abstract

“…Indeed, arranging systematically huge photo collections and retrieving specific pictures from them can be a daunting task, which becomes more and more difficult as time passes by ( [1,2]). This has initiated extensive research on content-based image retrieval systems ( [3,4,5,6,7,8,9]). Digital photographs typically include metadata in a standard image header, such as time, date and Global Positioning System (GPS) information that can be used for automatic organization.…”

Section: Introductionmentioning

confidence: 99%

Smartphone picture organization: A hierarchical approach

Lönn

Radeva

Dimiccoli

2019

Computer Vision and Image Understanding

View full text Add to dashboard Cite

We live in a society where the large majority of the population has a camera-equipped smartphone. In addition, hard drives and cloud storage are getting cheaper and cheaper, leading to a tremendous growth in stored personal photos. Unlike photo collections captured by a digital camera, which typically are pre-processed by the user who organizes them into event-related folders, smartphone pictures are automatically stored in the cloud. As a consequence, photo collections captured by a smartphone are highly unstructured and because smartphones are ubiquitous, they present a larger variability compared to pictures captured by a digital camera. To solve the need of organizing large smartphone photo collections automatically, we propose here a new methodology for hierarchical photo organization into topics and topic-related categories. Our approach successfully estimates latent topics in the pictures by applying probabilistic Latent Semantic Analysis, and automatically assigns a name to each topic by relying on a lexical database. Topic-related categories are then estimated by using a set of topic-specific Convolutional Neuronal Networks. To validate our approach, we ensemble and make public a large dataset of more than 8,000 smartphone pictures from 40 persons. Experimental results demonstrate major user satisfaction with respect to state of the art solutions in terms of organization.

show abstract

“…Geotagging also has grown in popularity as text messaging systems, social media outlets, and photo sharing sites (in particular Flickr.com) have enabled users to include geographic information with these various media (Barve 2014;Kumar and Seitz 2014). Participation in, and demand for, this functionality illustrates a general public interest in working with geographic interfaces, expanding geographic data and improving freely available geographic information.…”

mentioning

confidence: 99%

Mapping Life – Quality Assessment of Novice vs. Expert Georeferencers

Ellwood¹,

Bart²,

Doosey³

et al. 2016

CSTP

View full text Add to dashboard Cite

The majority of the world's billions of biodiversity specimens are tucked away in museum cabinets with only minimal, if any, digital records of the information they contain. Global efforts to digitize specimens are underway, yet the scale of the task is daunting. Fortunately, many activities associated with digitization do not require extensive training and could benefit from the involvement of citizen science participants. However, the quality of the data generated in this way is not well understood. With two experiments presented here, we examine the efficacy of citizen science participants in georeferencing specimen collection localities. In the absence of an online citizen science georeferencing platform and community, students served as a proxy for the larger citizen science population. At Tulane University and Florida State University, undergraduate students and experts used the GEOLocate platform to georeference fish and plant specimen localities, respectively. Our results provide a first-approximation of what can be expected from citizen science participants with minimal georeferencing training as a benchmark for future innovations. After outliers were removed, the range between student and expert georeferenced points was <1.0 to ca. 40.0 km for both the fish and the plant experiments, with an overall mean of 8.3 km and 4.4 km, respectively. Engaging students in the process improved results beyond GEOLocate's algorithm alone. Calculation of a median point from replicate points improved results further, as did recognition of good georeferencers (e.g., creation of median points contributed by the best 50% of contributors). We provide recommendations for improving accuracy further. We call for the creation of an online citizen science georeferencing platform.

show abstract

Photo Recall: Using the Internet to Label Your Photos

Cited by 10 publications

References 23 publications

Set2Model Networks: Learning Discriminatively To Learn Generative Models

Set2Model Networks: Learning Discriminatively To Learn Generative Models

Smartphone picture organization: A hierarchical approach

Mapping Life – Quality Assessment of Novice vs. Expert Georeferencers

Contact Info

Product

Resources

About