{"id":18487945,"url":"https://github.com/srama2512/sidekicks","last_synced_at":"2025-04-08T20:32:08.047Z","repository":{"id":33857166,"uuid":"140138546","full_name":"srama2512/sidekicks","owner":"srama2512","description":"Sidekick Policy Learning for Active Visual Exploration (ECCV 2018)","archived":false,"fork":false,"pushed_at":"2022-01-21T19:42:38.000Z","size":92,"stargazers_count":26,"open_issues_count":1,"forks_count":4,"subscribers_count":1,"default_branch":"master","last_synced_at":"2025-03-23T17:51:28.344Z","etag":null,"topics":["exploration","learning","reinforcement","visual"],"latest_commit_sha":null,"homepage":null,"language":"Python","has_issues":true,"has_wiki":null,"has_pages":null,"mirror_url":null,"source_name":null,"license":"mit","status":null,"scm":"git","pull_requests_enabled":true,"icon_url":"https://github.com/srama2512.png","metadata":{"files":{"readme":"README.md","changelog":null,"contributing":null,"funding":null,"license":"LICENSE","code_of_conduct":null,"threat_model":null,"audit":null,"citation":null,"codeowners":null,"security":null,"support":null}},"created_at":"2018-07-08T04:56:13.000Z","updated_at":"2024-10-22T09:19:43.000Z","dependencies_parsed_at":"2022-08-07T23:16:56.844Z","dependency_job_id":null,"html_url":"https://github.com/srama2512/sidekicks","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/srama2512%2Fsidekicks","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/srama2512%2Fsidekicks/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/srama2512%2Fsidekicks/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/srama2512%2Fsidekicks/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/srama2512","download_url":"https://codeload.github.com/srama2512/sidekicks/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":247923180,"owners_count":21018943,"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":["exploration","learning","reinforcement","visual"],"created_at":"2024-11-06T12:51:00.942Z","updated_at":"2025-04-08T20:32:03.037Z","avatar_url":"https://github.com/srama2512.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# Emergence of exploratory look-around behaviors through active observation completion\nA journal version of this work in conjunction with our prior work on [Learning to Look Around: Intelligently Exploring Unseen Environments for Unknown Tasks](https://arxiv.org/abs/1709.00507) has been published in Science Robotics 2019.\n\n[Emergence of exploratory look-around behaviors through active observation completion](http://vision.cs.utexas.edu/projects/visual-exploration/)  \nSanthosh K. Ramakrishnan, Dinesh Jayaraman, Kristen Grauman   \nScience Robotics 2019\n\nA cleaned version of this codebase along with new transfer tasks are available at https://github.com/srama2512/visual-exploration. \n\n# Sidekick Policy Learning\nThis repository contains code and data for the paper \n\n[Sidekick Policy Learning for Active Visual Exploration](http://vision.cs.utexas.edu/projects/sidekicks/)  \nSanthosh K. Ramakrishnan, Kristen Grauman  \nECCV 2018\n\n\n## Setup\n- First install anaconda and setup a new environment. Install anaconda from: https://www.anaconda.com/download/\n\n```\nconda create -n spl python=2.7\nsource activate spl\n```\n- Clone this project repository and setup requirements using pip.\n\n```\ngit clone https://github.com/srama2512/sidekicks.git\ncd sidekicks\npip install -r requirements.txt\n```\n\n- Download preprocessed SUN360 and ModelNet data.\n\n```\nwget http://vision.cs.utexas.edu/projects/sidekicks/data.zip\nunzip data.zip\n```\n\n- Sidekick scores for `ours-rew`, `ours-demo`, `rnd-rewards` on both datasets have been provided [here](http://vision.cs.utexas.edu/projects/sidekicks/scores.zip). The `one-view` model used to generate them have also been provided. \n \n## Evaluating pre-trained models\nAll the pre-trained models have been provided [here](http://vision.cs.utexas.edu/projects/sidekicks/models.zip). To evaluate them, download them to the `models` directory. To reproduce results from the paper: \n\n```\nwget http://vision.cs.utexas.edu/projects/sidekicks/models.zip\nunzip models.zip\nsh evaluation_script_final.sh\n```\n\n### Evaluation examples\n- Evaluating SUN360 `one-view` baseline on the test data with `avg` metric:\n\n```\npython eval.py --h5_path data/sun360/sun360_processed.h5 --dataset 0 \\\n\t\t\t\t  --model_path models/sun360/one-view.net --T 1 --M 8 --N 4 \\\n\t\t\t\t  --start_view 2 --save_path dummy/ \n```\n\n- Evaluating SUN360 `ltla` baseline on the test data with `avg` metric:\n\n```\npython eval.py --h5_path data/sun360/sun360_processed.h5 --dataset 0 \\\n\t\t\t\t  --model_path models/sun360/ltla.net --T 4 --M 8 --N 4 \\\n\t\t\t\t  --start_view 2 --save_path dummy/ \n```\n- Evaluating SUN360 `ltla` baseline on the test data with `adv` metric:\n\n```\npython eval.py --h5_path data/sun360/sun360_processed.h5 --dataset 0 \\\n\t\t\t\t  --model_path models/sun360/ltla.net --T 4 --M 8 --N 4 \\\n\t\t\t\t  --start_view 2 --save_path dummy/ \n```\n- Evaluating SUN360 `rnd-actions` baseline on test data with `avg` metric:\n\n```\npython eval.py --h5_path data/sun360/sun360_processed.h5 --dataset 0 \\\n\t\t\t\t  --model_path models/sun360/rnd-actions.net --T 4 --M 8 --N 4 \\\n\t\t\t\t  --start_view 2 --actorType random --save_path dummy/ \n```\n- Evaluating ModelNet Hard `one-view` baseline on test (seen and unseen) data with `avg` metric:\n\n```\npython eval.py --h5_path modelnet30_processed.h5 \\\n\t\t\t\t  --h5_path_unseen modelnet10_processed.h5 --dataset 1 \\\n\t\t\t\t  --model_path models/modelnet_hard/one-view.net --T 1 --M 9 --N 5 \\\n\t\t\t\t  --start_view 2 --save_path dummy/\n```\n\n## Training models\t\nEnsure that the [pre-trained models](http://vision.cs.utexas.edu/projects/sidekicks/models.zip) and [pre-computed scores](http://vision.cs.utexas.edu/projects/sidekicks/scores.zip) are downloaded and extracted. \n\n- Training `one-view` model on SUN360 with default settings:\n\n```\npython main.py --T 1 --training_setting 0 --epochs 100 \\\n\t\t\t\t  --save_path saved_models/sun360/one-view\n```\n- Training `ltla` baseline on SUN360 with default settings (starting from pre-trained `one-view` model): \n\n```\npython main.py --T 4 --training_setting 1 --epochs 1000 \\\n\t\t\t\t  --save_path saved_models/sun360/ltla/  \\\n\t\t\t\t  --load_model models/sun360/one-view.net\n```\n- Training `ours-rew` on SUN360 with default settings (with pre-computed score):\n\n```\npython main.py --T 4 --training_setting 1 --epochs 1000 \\\n\t\t\t\t  --save_path saved_models/sun360/ours-rew/ \\\n\t\t\t\t  --load_model models/sun360/one-view.net --expert_rewards True \\\n\t\t\t\t  --rewards_h5_path scores/sun360/ours-rew-scores.h5\n```\n- Training `ours-demo` on SUN360 with default settings (with pre-computed score):\n\n```\npython main.py --T 4 --training_setting 1 --epochs 1000 \\\n\t\t\t\t  --save_path saved_models/sun360/ours-demo/ \\\n\t\t\t\t  --load_model models/sun360/one-view.net --expert_trajectories True \\\n\t\t\t\t  --utility_h5_path scores/sun360/ours-demo-scores.h5\n```\n- Training `ltla` baseline on ModelNet Hard with default settings (starting from pre-trained `one-view` model):\n\n```\npython main.py --h5_path data/modelnet_hard/modelnet30_processed.h5 \\\n\t\t\t\t  --training_setting 1 --dataset 1 --T 4 --M 9 --N 5 \\\n\t\t\t\t  --load_model models/modelnet_hard/one-view.net \\\n\t\t\t\t  --save_path saved_models/modelnet_hard/ltla/\n```\n\nThe other ModelNet Hard models can be trained similar to SUN360 models. To train actor critic models, set `--baselineType critic`. To add full observability to the critic (for `asymm-ac`), set `--critic_full_obs True`. \n\n## Visualization\nFrom the repository directory, start jupyter notebook and open `visualize_policy_paper.ipynb`. Perform the TODOs mentioned in the comments (setting the correct paths) and run the entire script. It will generate tensorboard files contained visualized heatmaps on several examples. \n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsrama2512%2Fsidekicks","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fsrama2512%2Fsidekicks","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fsrama2512%2Fsidekicks/lists"}