FGVC launches Computer Vision Challenge for Recognizing hotels to Combat Human Trafficking

Human Trafficking is a serious problem all over the world which involves involuntary and forced transportation of humans for various purposes. Solving the problem of human trafficking is as big a problem as human trafficking itself. To address this issue, Fine-Grained Visual Categorization(FGVC) has launched the Hotel-ID to Combat Human Trafficking 2022 – FGVC9 challenge on Kaggle.

Victims of human trafficking are often photographed in hotel rooms. While identifying these hotels can prove vital to combat these trafficking investigations, it poses some particularly difficult challenges due to low quality of images and uncommon camera angles.

Sample Image of a victim of Human Trafficking
An Image of a victim of human trafficking in a hotel room. Source

Even without victims in the images, hotel identification is a challenging fine-grained visual recognition task with a huge number of classes and potentially high intraclass and low interclass variation.

To support research into this challenging task and create image search tools for human trafficking investigators, TraffickCam mobile application was created to allow every day travellers to submit photos of their hotel room.

Example of Images of a hotel room from FGVC Human Trafficking dataset
Example of images from a hotel room in the TraffickCam dataset

The Competition

In this competition, your task will be to identify the hotel based on images of hotel rooms, i.e, performing a mapping from hotel images to hotels. The training dataset provided will contain images from hotels and their corresponding hotel ids. The model should be trained to identify the hotel from images of hotel rooms in the test dataset.

The images in the test dataset are all partially masked to emulate the added challenge of having a person in the blocking part of the view of the room. Models that can overcome this extra layer of difficulty should be even more useful in practice.

Evaluation

The solution will be evaluated by the Mean Average Precision @ 5 (MAP@5) metric.

Requirements

  • Use of pre-trained models for the competition is allowed.
  • The code submissions will be in the form of Notebooks. Both CPU and GPU notebooks should have a run time of less than 9 hours.

Timelines

The competition began on 16th March 2022. And while the final submission deadline is on 30th May 2022, the deadline for entry into the competition and team merger is 23rd May 2022. There have been no competitors or entries so far in this competition.

Prizes

There is no cash prize for this competition.

Since this competition is a part of the Fine-Grained Visual Categorization, FGVC9 workshop at the Computer Vision and Pattern Recognition Conference CVPR 2022, a panel will review the top submissions for the competition based on the description of the methods provided. From this, a subset may be invited to present their results at the workshop. Attending the workshop is not required to participate in the competition; however, only teams that are attending the workshop will be considered to present their work.

To be a part of the Hotel-ID to Combat Human Trafficking 2022 – FGVC 9 challenge, you just need to have a Kaggle Account. To know more about this challenge and participate in it, click here.

Good Luck!


Other Articles you might be Interested In


FGVC and GeoLifeCLEF launch Computer Vision Competition for predicting Location-based Species Presence

Ministry of Electronics and Information Technology launches TIDE 2.0 Challenge Grant for Emerging Technologies

Qiskit opens applications for Quantum Error Correction Summer School 2022

ConvNext: The ConvNet for the Roaring 20s

Getting Started with Quantum Computing? Check out these Resources!