Hello everyone!
\nWe're excited to launch the first Hotel-ID to Combat Human Trafficking challenge. This challenge is part of the Eighth Workshop on Fine-Grained Visual Categorization (FGVC8) at CVPR 2021.
\nFor the last several years, our team of researchers at Saint Louis University, George Washington University and Temple University has worked to build large scale datasets of hotel room imagery to combat human trafficking. You can read some about our efforts here.
\nVictims of human trafficking are often photographed in hotel rooms and recognizing those hotels is an important part of the investigations against traffickers -- for example, in the United States, if it can be proved that a victim was photographed in a hotel in two different states, then the trafficking charge becomes a federal, rather than state, crime.
\nHotel recognition is, however, quite challenging. Rooms within the same hotel can look very different, and rooms from different hotels (especially those within the same chain) might look quite similar. Not only that, but images in investigations often are taken on smart phones and look quite different from the sorts of images you might find in promotional materials from a hotel. To address that, we developed a mobile application called TraffickCam, that allows every day travelers to upload images of their hotel to combat trafficking.
\nIn this competition, we are releasing a dataset with these TraffickCam images from thousands of hotels, along with a test set of images from unknown (to you!) hotels. Your challenge is to come up with the best strategy to predict the hotel that is seen in the pictures in this test set.
\nOur team has worked over the last several years to train hotel recognition models using our datasets to support an image recognition tool that is deployed at the National Center for Missing and Exploited Children. We're excited to see the sorts of new approaches that you propose in this contest, and work with top performing competitors on possibly integrating solutions into this tool.
\nOur team is really looking forward to seeing the sorts of approaches that you come up with! We'll be around on Kaggle to answer questions that might come up.
\nThanks,
\nAbby Stylianou (Saint Louis University)
\nRashmi Kamath (Saint Louis University)
\nRichard Souvenir (Temple University)
\nRobert Pless (George Washington University)
\nand the rest of the TraffickCam team!
Hello everyone!
\nWe're excited to launch the first Hotel-ID to Combat Human Trafficking challenge. This challenge is part of the Eighth Workshop on Fine-Grained Visual Categorization (FGVC8) at CVPR 2021.
\nFor the last several years, our team of researchers at Saint Louis University, George Washington University and Temple University has worked to build large scale datasets of hotel room imagery to combat human trafficking. You can read some about our efforts here.
\nVictims of human trafficking are often photographed in hotel rooms and recognizing those hotels is an important part of the investigations against traffickers -- for example, in the United States, if it can be proved that a victim was photographed in a hotel in two different states, then the trafficking charge becomes a federal, rather than state, crime.
\nHotel recognition is, however, quite challenging. Rooms within the same hotel can look very different, and rooms from different hotels (especially those within the same chain) might look quite similar. Not only that, but images in investigations often are taken on smart phones and look quite different from the sorts of images you might find in promotional materials from a hotel. To address that, we developed a mobile application called TraffickCam, that allows every day travelers to upload images of their hotel to combat trafficking.
\nIn this competition, we are releasing a dataset with these TraffickCam images from thousands of hotels, along with a test set of images from unknown (to you!) hotels. Your challenge is to come up with the best strategy to predict the hotel that is seen in the pictures in this test set.
\nOur team has worked over the last several years to train hotel recognition models using our datasets to support an image recognition tool that is deployed at the National Center for Missing and Exploited Children. We're excited to see the sorts of new approaches that you propose in this contest, and work with top performing competitors on possibly integrating solutions into this tool.
\nOur team is really looking forward to seeing the sorts of approaches that you come up with! We'll be around on Kaggle to answer questions that might come up.
\nThanks,
\nAbby Stylianou (Saint Louis University)
\nRashmi Kamath (Saint Louis University)
\nRichard Souvenir (Temple University)
\nRobert Pless (George Washington University)
\nand the rest of the TraffickCam team!
@abbystylianou
\nThanks for hosting this interesting competition.
\nCould you share test score of your hotel recognition model?
\nI would like to know how good (or inferior) my method is.
I'm a bit confused about the comp rules.
\nA3 says:
\nThe general rule is that participants should only use the provided training images for training models to classify the test images. We do not want participants crawling the web in search of additional data.
\nBut B7C contains the general external data clause.
\nSo, are we allowed to use data such as Hotels50k dataset https://github.com/GWUvision/Hotels-50K?
", "rawMarkdown": "I'm a bit confused about the comp rules.\n\nA3 says:\n\nThe general rule is that participants should only use the provided training images for training models to classify the test images. We do not want participants crawling the web in search of additional data.\n\nBut B7C contains the general external data clause.\n\nSo, are we allowed to use data such as Hotels50k dataset https://github.com/GWUvision/Hotels-50K?", "votes": null } ], "comments": [ { "id": 1255394, "author_name": "phalanx", "author_url": "", "post_date": "03/28/2021 18:43:55", "content": "@abbystylianou
\nThanks for hosting this interesting competition.
\nCould you share test score of your hotel recognition model?
\nI would like to know how good (or inferior) my method is.
I'm a bit confused about the comp rules.
\nA3 says:
\nThe general rule is that participants should only use the provided training images for training models to classify the test images. We do not want participants crawling the web in search of additional data.
\nBut B7C contains the general external data clause.
\nSo, are we allowed to use data such as Hotels50k dataset https://github.com/GWUvision/Hotels-50K?
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1234643": "Hello everyone!\n\nWe're excited to launch the first Hotel-ID to Combat Human Trafficking challenge. This challenge is part of the [Eighth Workshop on Fine-Grained Visual Categorization (FGVC8)](https://sites.google.com/corp/view/fgvc8) at [CVPR 2021](http://cvpr2021.thecvf.com/).\n\nFor the last several years, our team of researchers at Saint Louis University, George Washington University and Temple University has worked to build large scale datasets of hotel room imagery to combat human trafficking. [You can read some about our efforts here.](https://techcrunch.com/2016/06/25/traffickcam/)\n\nVictims of human trafficking are often photographed in hotel rooms and recognizing those hotels is an important part of the investigations against traffickers -- for example, in the United States, if it can be proved that a victim was photographed in a hotel in two different states, then the trafficking charge becomes a federal, rather than state, crime.\n\nHotel recognition is, however, quite challenging. Rooms within the same hotel can look very different, and rooms from different hotels (especially those within the same chain) might look quite similar. Not only that, but images in investigations often are taken on smart phones and look quite different from the sorts of images you might find in promotional materials from a hotel. To address that, we developed a mobile application called TraffickCam, that allows every day travelers to upload images of their hotel to combat trafficking.\n\nIn this competition, we are releasing a dataset with these TraffickCam images from thousands of hotels, along with a test set of images from unknown (to you!) hotels. Your challenge is to come up with the best strategy to predict the hotel that is seen in the pictures in this test set.\n\nOur team has worked over the last several years to train hotel recognition models using our datasets to support an image recognition tool that is deployed at the National Center for Missing and Exploited Children. We're excited to see the sorts of new approaches that you propose in this contest, and work with top performing competitors on possibly integrating solutions into this tool.\n\nOur team is really looking forward to seeing the sorts of approaches that you come up with! We'll be around on Kaggle to answer questions that might come up.\n\nThanks,\n\nAbby Stylianou (Saint Louis University)\nRashmi Kamath (Saint Louis University)\nRichard Souvenir (Temple University)\nRobert Pless (George Washington University)\nand the rest of the TraffickCam team!", "1255394": "abbystylianou \nThanks for hosting this interesting competition.\nCould you share test score of your hotel recognition model?\nI would like to know how good (or inferior) my method is.", "1299784": "I'm a bit confused about the comp rules.\n\nA3 says:\n\nThe general rule is that participants should only use the provided training images for training models to classify the test images. We do not want participants crawling the web in search of additional data.\n\nBut B7C contains the general external data clause.\n\nSo, are we allowed to use data such as Hotels50k dataset https://github.com/GWUvision/Hotels-50K?" }, "source": "meta" }