{"id":28433964,"url":"https://github.com/szemenyeim/robocupvision","last_synced_at":"2026-05-14T21:33:09.219Z","repository":{"id":109009960,"uuid":"114206190","full_name":"szemenyeim/RoboCupVision","owner":"szemenyeim","description":"Deep Learning for Semantic Segmentation in RoboCup","archived":false,"fork":false,"pushed_at":"2022-05-18T06:39:32.000Z","size":23485,"stargazers_count":1,"open_issues_count":0,"forks_count":2,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-10-21T12:45:56.835Z","etag":null,"topics":["dataset","deep-learning","neural-network","pretrained-models","pytorch","pytorch-implementation","robocup","robocup-spl","segmentation-networks","semantic-segmentation"],"latest_commit_sha":null,"homepage":null,"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/szemenyeim.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,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null,"zenodo":null}},"created_at":"2017-12-14T05:24:03.000Z","updated_at":"2025-06-23T14:33:41.000Z","dependencies_parsed_at":"2023-03-23T13:33:21.569Z","dependency_job_id":null,"html_url":"https://github.com/szemenyeim/RoboCupVision","commit_stats":null,"previous_names":[],"tags_count":0,"template":false,"template_full_name":null,"purl":"pkg:github/szemenyeim/RoboCupVision","repository_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szemenyeim%2FRoboCupVision","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szemenyeim%2FRoboCupVision/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szemenyeim%2FRoboCupVision/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szemenyeim%2FRoboCupVision/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/szemenyeim","download_url":"https://codeload.github.com/szemenyeim/RoboCupVision/tar.gz/refs/heads/master","sbom_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/szemenyeim%2FRoboCupVision/sbom","scorecard":null,"host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":286080680,"owners_count":33044118,"icon_url":"https://github.com/github.png","version":null,"created_at":"2022-05-30T11:31:42.601Z","updated_at":"2026-05-13T13:14:54.681Z","status":"online","status_checked_at":"2026-05-14T02:00:06.663Z","response_time":57,"last_error":null,"robots_txt_status":"success","robots_txt_updated_at":"2025-07-24T06:49:26.215Z","robots_txt_url":"https://github.com/robots.txt","online":true,"can_crawl_api":true,"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":["dataset","deep-learning","neural-network","pretrained-models","pytorch","pytorch-implementation","robocup","robocup-spl","segmentation-networks","semantic-segmentation"],"created_at":"2025-06-05T18:31:03.885Z","updated_at":"2026-05-14T21:33:09.206Z","avatar_url":"https://github.com/szemenyeim.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# RoboCupVision\nDeep Learning for Semantic Segmentation in RoboCup\n\nThis set of scripts allows you to train deep convolutional networks for semantic segmentation in Pytorch as described in our paper. This code is finetuned to work well with the networks and the problem (robot soccer) we discussed in the paper, still it should work reasonably fine for other problems as well.\n\n## 1.Requirements\n\nThis code is written in python 3.6. The following python packages are required:\n\n- Pytorch\n- Torchvision\n- Progressbar 2\n- PIL Image library\n- Numpy\n- Visdom\n\nYou can install all of these by entering pip install \u0026lt;package name\u0026gt; into the console.\n\nIn order to run the code, you also need our datasets, which you can download here. Extract it to the data subfolder. (NOTE: If after extraction you have \u0026quot;data/data/many folders\u0026quot;, then you need to move the whole thing one folder up to get rid of one of the \u0026quot;data\u0026quot;-s.)\n\n## 2.Dataset and Pre-trained models\n\nYou can download the datasets used for training and finetuning from our website\n\n- [Train set](https://deeplearning.iit.bme.hu/Public/ROBOSeg/train.zip)\n- [Validation set](https://deeplearning.iit.bme.hu/Public/ROBOSeg/val.zip)\n- [Fintune set](https://deeplearning.iit.bme.hu/Public/ROBOSeg/FinetuneHorizon.zip)\n\nIf you do not want to train your own models, just use our pretrained models, which you can download from [here](https://deeplearning.iit.bme.hu/Public/ROBOSeg/checkpoints.zip). Extract them into the \u0026quot;checkpoints\u0026quot; folder.\n\n## 3.Training a segmentation network\n\nThe pretrain the whole segmentation network on synthetic data.\n\n        python train.py\n\nThen finetune the network on the real dataset, (you can use synthetic transfer learning optionally).\n\n        python train.py –-finetune\n\n        python train.py –-finetune --transfer\n\nThe finetune flag also determines if the synthetic or the real test sets are evaluated.\n\nThere are several further options you can use for training:\n\n- --noScale        This option will run the network on 640x480 resolution instead of 160x120\n- --UNet        This option will train the standard U-Net architecture, instead of the ROBO-UNet proposed in our paper.\n- --topCam, --bottomCam        Use top or bottom camera images only. Providing both or neither flags will use both cameras.\n- --noBall, --noRobot, --noGoal, --noLine        Treat balls, robot, goals or lines as background. You can use any combinations of these flags, as long as there is at least one foreground class.\n\nFinally, you can generate the labeled images using the tester script. You can use the flags to select different neural network versions. Examples:\n\n        python test.py\n\n        python test.py –-finetune\n\n        python tester.py –-finetune –-transfer\n\n        python tester.py –-finetune --v2 --topCam\n\n        python tester.py –-finetune --v2 --bottomCam --noGoal --noRobot\n\nThere are several further options you can use for training:\n\n## 4.Training the label propagation network\n\nThe training of the label propagation network is quite similar to the segmentation networks, except that the classification pretraining step is omitted.\n\n        python labelPropTrain.py\n\n        python labelPropTrain.py -–finetune\n\n        python labelPropTrain.py –-finetune –-prune\n\nYou can also generate the labeled images using the following script:\n\n        python validLabelProp.py –-finetuned –-pruned\n\nYou can also compare the results with farneback optical flow using:\n\n        python validLabelProp.py –-finetuned –-optFlow\n\nNote, that the \u0026#39;finetuned\u0026#39; flag is needed to set the database to the real one.\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fszemenyeim%2Frobocupvision","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fszemenyeim%2Frobocupvision","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fszemenyeim%2Frobocupvision/lists"}