{"id":19419262,"url":"https://github.com/mahdizynali/seglight","last_synced_at":"2025-06-24T15:42:06.522Z","repository":{"id":232824420,"uuid":"785278682","full_name":"mahdizynali/SegLight","owner":"mahdizynali","description":"Super fast and real-time semantic segmentation (cpu only) can be use for 1 core cpu","archived":false,"fork":false,"pushed_at":"2025-03-18T06:36:31.000Z","size":215,"stargazers_count":10,"open_issues_count":1,"forks_count":2,"subscribers_count":1,"default_branch":"main","last_synced_at":"2025-04-11T17:56:09.845Z","etag":null,"topics":["ai","deep-learning","image","keras-tensorflow","machine-learning","neural-network","object-detection","robot","robotics","segment-anything","semantic","semantic-segmentation","soccer-robots","tensorflow"],"latest_commit_sha":null,"homepage":"","language":"Python","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/mahdizynali.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":null,"code_of_conduct":null,"threat_model":null,"audit":null,"citation":"CITATION.cff","codeowners":null,"security":null,"support":null,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2024-04-11T15:05:46.000Z","updated_at":"2025-03-18T06:36:35.000Z","dependencies_parsed_at":"2024-04-21T01:46:31.132Z","dependency_job_id":"d5ee29e1-375d-49e3-86e5-90a81bc758eb","html_url":"https://github.com/mahdizynali/SegLight","commit_stats":null,"previous_names":["mahdizynali/seglight"],"tags_count":0,"template":false,"template_full_name":null,"repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mahdizynali%2FSegLight","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mahdizynali%2FSegLight/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mahdizynali%2FSegLight/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/mahdizynali%2FSegLight/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/mahdizynali","download_url":"https://codeload.github.com/mahdizynali/SegLight/tar.gz/refs/heads/main","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":250643421,"owners_count":21464172,"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":["ai","deep-learning","image","keras-tensorflow","machine-learning","neural-network","object-detection","robot","robotics","segment-anything","semantic","semantic-segmentation","soccer-robots","tensorflow"],"created_at":"2024-11-10T13:16:59.269Z","updated_at":"2025-04-24T14:31:42.974Z","avatar_url":"https://github.com/mahdizynali.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"## Light Semantic Segmentation\nApproch of this project is implementing light weight semantic segmentation model which has the least inference time as possible on cpu that we use on humanoid soccer robots. Also network architecture has been inspired from our last team project.\n## How to use\nfirst of all you have to prepare suitable semantic dataset as images and labels files in dataset directory like bellow :\n```\ndatase/\n  |_ seri1:\n    |_ /images/\n    |_ /labels/\n  |_ seri2:\n    |_ /images/\n    |_ /labels/\n  ...\n```\n![alt text](https://raw.githubusercontent.com/mahdizynali/SegLight/main/dataset/images/new_46.png)\n![alt text](https://github.com/mahdizynali/SegLight/blob/main/dataset/labels/new_46.png) \\\nThen you have to set your configuration in config file and intiate your semantic color-map.\n#### Notice : if you don't have reach dataset, you would use repeat option in augmentation data_provider file :\n```\ntrain_dataset = train_dataset.repeat(60)\nrepeat dataset 60 times !!\n```\n# Save Model Hint !\nin tensorflow version 2.16.0 and above, keras kernel updates into version 3 and it limit us to save models only in .h5 or .keras format;\nso as i wanna inference on cpp and cppflow, i need to save as tf format .pb as keras v2 in order to load in cppflow inferencer.\\\ntry to install also keras v2 :\n```\npip install tf-keras~=2.16\n```\nthen in directory of your project set env :\n```\nexport TF_USE_LEGACY_KERAS=1\n```\nin main code you would change formats as you wish :\n```\nmodel.save(\"save/path\", save_format='tf') # for keras v2\nmodel.save(\"model.h5\") # or may .keras for keras v3\n``` \nFinally try to run main.py as trainer file to store trained model into that specific folder which you set in main.py.\n## Citation\n```\n@software{Mahdi_SegLight_Light_Semantic,\n  author = {Mahdi, Zeinali},{Erfan, Ramezani},\n  title = {{SegLight (Light Semantic Segmentation For Humanoid Soccer Robots)}},\n  url = {https://github.com/mahdizynali/SegLight},\n  version = {1.0}\n}\n```\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmahdizynali%2Fseglight","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fmahdizynali%2Fseglight","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fmahdizynali%2Fseglight/lists"}