# Image Similarity Challenge **URL:** https://community.drivendata.org/c/image-similarity-challenge/46.md [Latest](https://community.drivendata.org/latest.md) · [Categories](https://community.drivendata.org/categories.md) --- ## [About the Image Similarity Challenge category](https://community.drivendata.org/t/about-the-image-similarity-challenge-category/6193)
**Author:** [@mike-dd](https://community.drivendata.org/u/mike-dd)\ **Replies:** 0
This category is for posts about either track of the Facebook AI Image Similarity Challenge: Matching Track Descriptor Track Please keep all posts here specific to this competition. --- ## [Training code for GeM baseline](https://community.drivendata.org/t/training-code-for-gem-baseline/6271)
**Author:** [@LexiBender](https://community.drivendata.org/u/LexiBender)\ **Replies:** 4\ **Last updated:** [August 30, 2022, 10:33am UTC](https://community.drivendata.org/t/training-code-for-gem-baseline/6271 "2022-08-30T10:33:59Z")
Where is a training code to train GeM baseline? https://github.com/facebookresearch/isc2021/blob/master/baselines/GeM\_baseline.py --- ## [Will isc dataset provide labels in the future?](https://community.drivendata.org/t/will-isc-dataset-provide-labels-in-the-future/6711)
**Author:** [@ZaneRo](https://community.drivendata.org/u/ZaneRo)\ **Replies:** 8\ **Last updated:** [June 7, 2022, 1:48am UTC](https://community.drivendata.org/t/will-isc-dataset-provide-labels-in-the-future/6711 "2022-06-07T01:48:40Z")
Hi, @dd-mike ，We want to do some experiments on ISC data sets. Will their labels be published in the future？ --- ## [Release of Winning Solutions](https://community.drivendata.org/t/release-of-winning-solutions/6737)
**Author:** [@avishek27](https://community.drivendata.org/u/avishek27)\ **Replies:** 1\ **Last updated:** [November 22, 2021, 2:15pm UTC](https://community.drivendata.org/t/release-of-winning-solutions/6737 "2021-11-22T14:15:21Z")
Will the winning solutions be released to public? I am interested to learn about the techniques/architectures used by winners and their winning method. --- ## [Submission to Phase 2 without submitting to End Phase 1](https://community.drivendata.org/t/submission-to-phase-2-without-submitting-to-end-phase-1/6706)
**Author:** [@gkordo](https://community.drivendata.org/u/gkordo)\ **Replies:** 2\ **Last updated:** [October 27, 2021, 12:49pm UTC](https://community.drivendata.org/t/submission-to-phase-2-without-submitting-to-end-phase-1/6706 "2021-10-27T12:49:11Z")
Hi @dd-mike, I have not made a submission to the End Phase 1 of Descriptor Track. Still, I am interested in submitting my extracted features for evaluation in Phase 2 (certainly, out of competition) just to get their per… --- ## [One question about Phase 2](https://community.drivendata.org/t/one-question-about-phase-2/6697)
**Author:** [@wenhaowang](https://community.drivendata.org/u/wenhaowang)\ **Replies:** 4\ **Last updated:** [October 25, 2021, 11:53pm UTC](https://community.drivendata.org/t/one-question-about-phase-2/6697 "2021-10-25T23:53:14Z")
Whether the script to submit the track2 h5 file is the same as phase 1, except for the query IDs. That means it will be something like this: import h5py import numpy as np from isc.io import read\_descriptors, write\_hdf5… --- ## [Regarding Instructions for downloading data for phase2](https://community.drivendata.org/t/regarding-instructions-for-downloading-data-for-phase2/6696)
**Author:** [@AishwaryaH](https://community.drivendata.org/u/AishwaryaH)\ **Replies:** 3\ **Last updated:** [October 22, 2021, 9:41pm UTC](https://community.drivendata.org/t/regarding-instructions-for-downloading-data-for-phase2/6696 "2021-10-22T21:41:25Z")
Not able to find the instructions for downloading data for phase 2 . --- ## [Questions about End Phase 1 Submission](https://community.drivendata.org/t/questions-about-end-phase-1-submission/6683)
**Author:** [@lyakaap](https://community.drivendata.org/u/lyakaap)\ **Replies:** 5\ **Last updated:** [October 22, 2021, 3:48pm UTC](https://community.drivendata.org/t/questions-about-end-phase-1-submission/6683 "2021-10-22T15:48:21Z")
End Phase 1 Submission page says about h5 file like this: “This will be the finalized Phase 1 version of the file that you’ve already been submitting for evaluation.”, but I would like to submit other version which hasn’… --- ## [Cannot complete the End Phase 1 submission](https://community.drivendata.org/t/cannot-complete-the-end-phase-1-submission/6686)
**Author:** [@dragon6](https://community.drivendata.org/u/dragon6)\ **Replies:** 3\ **Last updated:** [October 19, 2021, 7:24pm UTC](https://community.drivendata.org/t/cannot-complete-the-end-phase-1-submission/6686 "2021-10-19T19:24:21Z")
@dd-mike I have been trying to submit the End Phase 1 submission but the webpage seems to be stuck forever. Is this a known issue? Do you have any other submission methods? Many thanks! --- ## [The validation of a submission](https://community.drivendata.org/t/the-validation-of-a-submission/6682)
**Author:** [@wenhaowang](https://community.drivendata.org/u/wenhaowang)\ **Replies:** 3\ **Last updated:** [October 19, 2021, 5:23pm UTC](https://community.drivendata.org/t/the-validation-of-a-submission/6682 "2021-10-19T17:23:01Z")
@dd-mike Our team has submitted two zip files to two tracks. However, we cannot download them after submitting them. Therefore, we want to confirm our submissions are valid. Thanks! --- ## [Benchmark Model](https://community.drivendata.org/t/benchmark-model/6239)
**Author:** [@jimking100](https://community.drivendata.org/u/jimking100)\ **Replies:** 3\ **Last updated:** [October 19, 2021, 5:20pm UTC](https://community.drivendata.org/t/benchmark-model/6239 "2021-10-19T17:20:43Z")
Hi, Are the details for the Benchmark Model (naive GIST descriptors) mentioned in the Performance Metric available? --- ## [What is the use of Similarity normalization](https://community.drivendata.org/t/what-is-the-use-of-similarity-normalization/6669)
**Author:** [@ZaneRo](https://community.drivendata.org/u/ZaneRo)\ **Replies:** 6\ **Last updated:** [October 19, 2021, 1:37am UTC](https://community.drivendata.org/t/what-is-the-use-of-similarity-normalization/6669 "2021-10-19T01:37:53Z")
Hi, I don’t understand why use similarity normalization. If the Euclidean distance of the descriptor is used to measure the similarity, then the similarity is independent of other reference images.So we don’t need simi… --- ## [External data (flickr)](https://community.drivendata.org/t/external-data-flickr/6676)
**Author:** [@LexiBender](https://community.drivendata.org/u/LexiBender)\ **Replies:** 1\ **Last updated:** [October 18, 2021, 5:51pm UTC](https://community.drivendata.org/t/external-data-flickr/6676 "2021-10-18T17:51:33Z")
We used a small subset of the FLICKR dataset to use it for augmentations (images overlaying and so on). Is it okay in terms of rules? --- ## [The format to submit Track 2 reference features](https://community.drivendata.org/t/the-format-to-submit-track-2-reference-features/6670)
**Author:** [@wenhaowang](https://community.drivendata.org/u/wenhaowang)\ **Replies:** 4\ **Last updated:** [October 15, 2021, 3:24pm UTC](https://community.drivendata.org/t/the-format-to-submit-track-2-reference-features/6670 "2021-10-15T15:24:56Z")
@dd-mike Hi, what is the format to submit the reference features of Track 2? It is similar to the given with h5py.File(out, "w") as f: f.create\_dataset("query", data=M\_query) f.create\_dataset("reference", data=M… --- ## [Do I need to submit model weights?](https://community.drivendata.org/t/do-i-need-to-submit-model-weights/6665)
**Author:** [@separate](https://community.drivendata.org/u/separate)\ **Replies:** 4\ **Last updated:** [October 15, 2021, 6:53pm UTC](https://community.drivendata.org/t/do-i-need-to-submit-model-weights/6665 "2021-10-15T18:53:50Z")
Hi @dd-mike, I’d like to know whether I need to submit model weights for phase 2 submission. Thank you. --- ## [Confusion on what to train the model on](https://community.drivendata.org/t/confusion-on-what-to-train-the-model-on/6662)
**Author:** [@NawasNaziru](https://community.drivendata.org/u/NawasNaziru)\ **Replies:** 1\ **Last updated:** [October 13, 2021, 5:24pm UTC](https://community.drivendata.org/t/confusion-on-what-to-train-the-model-on/6662 "2021-10-13T17:24:58Z")
If it is not recommended to train on the reference data, then as per the model output requirement which needs the reference id, how can one connect the training data to the reference data since the model doesn’t know any… --- ## [Query regarding deadline dates for Phase 1](https://community.drivendata.org/t/query-regarding-deadline-dates-for-phase-1/6637)
**Author:** [@sneezygiraffe](https://community.drivendata.org/u/sneezygiraffe)\ **Replies:** 3\ **Last updated:** [October 13, 2021, 5:20pm UTC](https://community.drivendata.org/t/query-regarding-deadline-dates-for-phase-1/6637 "2021-10-13T17:20:35Z")
Hi team, I wanted to confirm the submission deadlines for phase 1. I did not find an exact date for the same (probably I missed it). The main page says - “Phase 1: Model Development (June - October 2021)”, " Phase 2: Fi… --- ## [Which Deep Learning Model can we use?](https://community.drivendata.org/t/which-deep-learning-model-can-we-use/6666)
**Author:** [@AbhishekBhattad](https://community.drivendata.org/u/AbhishekBhattad)\ **Replies:** 0\ **Last updated:** [October 13, 2021, 9:28am UTC](https://community.drivendata.org/t/which-deep-learning-model-can-we-use/6666 "2021-10-13T09:28:52Z")
Hi Team, Can you guys help with which deep learning model is used. Its for educational purpose, we have this competition as a course project in Data Mining and analytics course. We have found deep ranking model is good… --- ## [Labels for training data](https://community.drivendata.org/t/labels-for-training-data/6652)
**Author:** [@NawasNaziru](https://community.drivendata.org/u/NawasNaziru)\ **Replies:** 0\ **Last updated:** [October 9, 2021, 9:14am UTC](https://community.drivendata.org/t/labels-for-training-data/6652 "2021-10-09T09:14:28Z")
How do I get the labels for the training data such that they correspond to the reference data? --- ## [Is augmenting reference images for local validation allowed?](https://community.drivendata.org/t/is-augmenting-reference-images-for-local-validation-allowed/6612)
**Author:** [@separate](https://community.drivendata.org/u/separate)\ **Replies:** 7\ **Last updated:** [October 8, 2021, 12:12am UTC](https://community.drivendata.org/t/is-augmenting-reference-images-for-local-validation-allowed/6612 "2021-10-08T00:12:18Z")
Hi, I’d like to know if augmenting reference images for local validation purpose is allowed in this competition. Thank you. --- ## [Is 0.4991 a pretrained model?](https://community.drivendata.org/t/is-0-4991-a-pretrained-model/6315)
**Author:** [@hulu-cheng](https://community.drivendata.org/u/hulu-cheng)\ **Replies:** 2\ **Last updated:** [October 7, 2021, 9:37am UTC](https://community.drivendata.org/t/is-0-4991-a-pretrained-model/6315 "2021-10-07T09:37:39Z")
the leadeboard shows some teams having the same score of 0.4991, is there a pretained model? --- ## [Why is the use of augmented reference images prohibited?](https://community.drivendata.org/t/why-is-the-use-of-augmented-reference-images-prohibited/6642)
**Author:** [@ZaneRo](https://community.drivendata.org/u/ZaneRo)\ **Replies:** 0\ **Last updated:** [October 6, 2021, 6:05pm UTC](https://community.drivendata.org/t/why-is-the-use-of-augmented-reference-images-prohibited/6642 "2021-10-06T18:05:05Z")
Hi, I’ve read the rules, but I don’t quite understand this one: Use of augmented reference images for any other reason, including model training, is prohibited. Is it to avoid overfitting, or something else? --- ## [Question Regarding Phase 2](https://community.drivendata.org/t/question-regarding-phase-2/6626)
**Author:** [@separate](https://community.drivendata.org/u/separate)\ **Replies:** 2\ **Last updated:** [October 1, 2021, 4:48am UTC](https://community.drivendata.org/t/question-regarding-phase-2/6626 "2021-10-01T04:48:35Z")
Hi, I have 2 questions regarding phase 2. Whether if code submission must include training code. (not only inference) If I can use 3 different inference models for 3 submissions. Thank you. --- ## [Training on Reference images](https://community.drivendata.org/t/training-on-reference-images/6367)
**Author:** [@gradient.machine](https://community.drivendata.org/u/gradient.machine)\ **Replies:** 19\ **Last updated:** [September 30, 2021, 3:41pm UTC](https://community.drivendata.org/t/training-on-reference-images/6367 "2021-09-30T15:41:33Z")
I’m a bit confused that can I train on the reference set ? Since we need to compute the similarity between the reference and query images pair, it is necessary to learn the embeddings of the reference set. --- ## [Image Similarity Challenge - one month until Phase 2!](https://community.drivendata.org/t/image-similarity-challenge-one-month-until-phase-2/6609)
**Author:** [@mike-dd](https://community.drivendata.org/u/mike-dd)\ **Replies:** 3\ **Last updated:** [September 30, 2021, 3:37pm UTC](https://community.drivendata.org/t/image-similarity-challenge-one-month-until-phase-2/6609 "2021-09-30T15:37:33Z")
Greetings, Facebook Image Similarity Challenge participants! To everyone participating in the Image Similarity Challenge, thanks for all your great work so far! As a reminder, Phase 2 of the challenge starts in around … --- ## [Were can I download the dataset besides AWS?](https://community.drivendata.org/t/were-can-i-download-the-dataset-besides-aws/6585)
**Author:** [@ZaneRo](https://community.drivendata.org/u/ZaneRo)\ **Replies:** 3\ **Last updated:** [September 28, 2021, 7:25am UTC](https://community.drivendata.org/t/were-can-i-download-the-dataset-besides-aws/6585 "2021-09-28T07:25:15Z")
Hi, I don’t have an AWS account, because I don’t have a qualified credit or debit card, so I can’t download the data set.Can I download it somewhere else? --- ## [Is legal for training on the reference dataset?](https://community.drivendata.org/t/is-legal-for-training-on-the-reference-dataset/6293)
**Author:** [@wenhaowang](https://community.drivendata.org/u/wenhaowang)\ **Replies:** 1\ **Last updated:** [July 29, 2021, 7:42pm UTC](https://community.drivendata.org/t/is-legal-for-training-on-the-reference-dataset/6293 "2021-07-29T19:42:55Z")
Hi, thanks for the excellent challenge held by you. I want to know whether we can train models on the reference dataset rather than the training dataset under the rules. Thanks. Or whether we can train on the source of… --- ## [Score Normalization](https://community.drivendata.org/t/score-normalization/6558)
**Author:** [@greenfieldvision](https://community.drivendata.org/u/greenfieldvision)\ **Replies:** 2\ **Last updated:** [September 20, 2021, 4:10pm UTC](https://community.drivendata.org/t/score-normalization/6558 "2021-09-20T16:10:20Z")
Hi, I’m late to the party, but what is the official position regarding score normalization for track 2? The score normalization script (https://github.com/facebookresearch/isc2021/blob/main/scripts/score\_normalization.p… --- ## [Questions about Similarity Challenge Rules](https://community.drivendata.org/t/questions-about-similarity-challenge-rules/6546)
**Author:** [@johnnyg6809](https://community.drivendata.org/u/johnnyg6809)\ **Replies:** 2\ **Last updated:** [September 20, 2021, 2:45pm UTC](https://community.drivendata.org/t/questions-about-similarity-challenge-rules/6546 "2021-09-20T14:45:29Z")
Hi, I just wanted to clarify some of the rules for the Similarity Challenge: Are you allowed to set some of the data from the Training Set aside for Validation and Test Sets during Phase I of the challenge? and Gr… --- ## [Phase 2 Computation Time](https://community.drivendata.org/t/phase-2-computation-time/6591)
**Author:** [@jwhart1](https://community.drivendata.org/u/jwhart1)\ **Replies:** 0\ **Last updated:** [September 18, 2021, 10:40am UTC](https://community.drivendata.org/t/phase-2-computation-time/6591 "2021-09-18T10:40:53Z")
This competition seems to have an implicit requirement for a combination of low computation operations per inference and/or high FLOPS (which is not to say that the models developed for the competition are required to us… [Next page](https://community.drivendata.org/c/image-similarity-challenge/46.md?page=1)