{"id":19152483,"url":"https://github.com/gesiscss/improved-gender-identifier","last_synced_at":"2026-03-04T13:01:45.842Z","repository":{"id":56453287,"uuid":"178152057","full_name":"gesiscss/Improved-Gender-Identifier","owner":"gesiscss","description":"GIGIDL is a deep learning system for gender inference was trained on a massive IMDB and WIKI dataset.","archived":false,"fork":false,"pushed_at":"2020-11-27T16:07:21.000Z","size":4485,"stargazers_count":4,"open_issues_count":1,"forks_count":5,"subscribers_count":11,"default_branch":"master","last_synced_at":"2025-03-31T06:51:11.924Z","etag":null,"topics":["artificial-neural-networks","deep-learning","gender-inference","python"],"latest_commit_sha":null,"homepage":"http://193.175.238.89/Gender_Inference/","language":"Jupyter Notebook","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":null,"status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/gesiscss.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2019-03-28T07:43:59.000Z","updated_at":"2024-01-19T10:26:12.000Z","dependencies_parsed_at":"2022-08-15T19:00:59.249Z","dependency_job_id":null,"html_url":"https://github.com/gesiscss/Improved-Gender-Identifier","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gesiscss%2FImproved-Gender-Identifier","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gesiscss%2FImproved-Gender-Identifier/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gesiscss%2FImproved-Gender-Identifier/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/gesiscss%2FImproved-Gender-Identifier/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/gesiscss","download_url":"https://codeload.github.com/gesiscss/Improved-Gender-Identifier/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":249738800,"owners_count":21318504,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2022-07-04T15:15:14.044Z","host_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub","repositories_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories","repository_names_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repository_names","owners_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners"}},"keywords":["artificial-neural-networks","deep-learning","gender-inference","python"],"created_at":"2024-11-09T08:18:04.152Z","updated_at":"2026-03-04T13:01:40.814Z","avatar_url":"https://github.com/gesiscss.png","language":"Jupyter Notebook","funding_links":[],"categories":[],"sub_categories":[],"readme":"![alt text](http://193.175.238.89/Gender_Inference/static/images/logo_banner_v3.png)\n\n\n\n# GESIS Improved Gender Inference Deep Learning\n\n\n\n## Requirements\n\n* Keras (Using TensorFlow backend)\n\n## About\n\nGIGIDL takes vision inputs specifically image of a person or web address of a image. Generally GIGIDL identify faces from the image and take this single or multiple faces as a input and provides the gender of that face or faces with the probability of accuracy if it’s human otherwise it skip the prediction of image data.\n\n\n## Training\n\nTo train our model we have used full dataset of IMDB and WIKI together (after cleaning total 36,356 images of  human). We have used 10-fold cross-validation to minimize the bias and got 98.28 % training accuracy.\n\n\n\n## How to Use: \n\nThis model takes an input image and gives the prediction as an output. You need to \n        a) Copy or clone this package (```git clone https://github.com/gesiscss/Improved-Gender-Identifier.git```) and follow the instructions.\n        b) Preprocess (for a bunch of images) like resize to get them reshaped. Use the preprocess script to make the images right.\n\n\nYou can use this model to predict the gender using image or a bunch of images (same extension e.g., only jpg or png and so on ) as data in any format but need to mention the extension (e.g., jpg, png etc.). \n\nImage could be inputed mostly 2 different ways:\n\n   * Web-link\n\n   Simply provide the web address of the image or images as web link in the system (e.g., ```web_link = “https://github.com/gesiscss/Improved-Gender -Identifier/blob/master/GIGIDL/Data/a.jpeg”```).\n   \n   * Directory\n\n\n   You need to create a directory called ```‘data’``` in your default directory or individually and keep your image or images in that directory in the system (e.g., ```[default directory] /data/```).\n\n\n## Install\n\n\n\n\n\n\n## Datasets\n\n\n* [IMDB Dataset](https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/)\n\n\n\nWhat is in the data: The IMDB dataset contains 460,723 facial images (with gender and age labels) of film stars, predominantly Hollywood actors and actresses. We have taken only the frontal faces which has the face score 4.5 or above and get 33,147 facial images.\n\nNumber of images:  33,147\n\nMale images: 14370\n\nFemale images: 18777\n\n\n\n* [Wikipedia Dataset](https://data.vision.ee.ethz.ch/cvl/rrothe/imdb-wiki/)\n\n\nWhat is in the data: Wikipedia dataset includes 62,328 (with gender and age labels) of celebrities from various fields, such as sports, politics, social events, and the film industry. We have taken only the frontal faces which has the face score 5 or above and get 3,209 facial images.\n\nNumber of images: 3,209\n\nMale images: 1830\n\nFemale images: 1379\n\n\n\n* [Scholars Dataset (manual labeled at GESIS)](https://gesis.org/)\n\n\nWhat is in the data: Images of xx scholars which have been manually labeled by the coders at GESIS. \n\nNumber of images: 3,324\n\nMale images: 1,953\n\nFemale images: 1,371\n\n\n* Every performance have huge differences between male and female images. We have checked the reasons and got some problems which need to be minimized to get better performance for female and that is, \n\n        a) Bad performance for coloured data \n        b) Better performance when the canvas is not coloured i.e. noisy \n        c) More better when the canvas is perfectly white.\n\nNote: Quality images means low resolution, background is noisy, face is not frontal, side view or multiple objects in the image, have eye glass, cap or cloths top to the head and so on.\n\n\n\n## Limitations\n\nWe have some limitations in our training model i.e.,\n\n        a) Accuracy for females is lower than the accuracy for males.\n        b) Predictions for dark, low resolutions and less contrast images is lower.\n        c) For the people of east asia prediction accuracy is also lower.\n\n\n\n## FAQs\n\nIf you have any questions then contact with the team members.\n\n\n## More Questions\n\n## License\n\n\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgesiscss%2Fimproved-gender-identifier","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fgesiscss%2Fimproved-gender-identifier","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fgesiscss%2Fimproved-gender-identifier/lists"}