{"id":17358629,"url":"https://github.com/prouast/ctc-intake-detection","last_synced_at":"2026-05-04T02:35:57.553Z","repository":{"id":129543413,"uuid":"184370452","full_name":"prouast/ctc-intake-detection","owner":"prouast","description":"Automatic detection of intake gestures with CTC from inertial or video data","archived":false,"fork":false,"pushed_at":"2020-11-03T02:24:46.000Z","size":88,"stargazers_count":0,"open_issues_count":0,"forks_count":0,"subscribers_count":2,"default_branch":"master","last_synced_at":"2025-03-25T17:51:39.992Z","etag":null,"topics":["ctc","deep-learning","intake-gesture-detection","oreba-dataset","tensorflow"],"latest_commit_sha":null,"homepage":"","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/prouast.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,"governance":null,"roadmap":null,"authors":null,"dei":null,"publiccode":null,"codemeta":null}},"created_at":"2019-05-01T04:53:07.000Z","updated_at":"2020-11-03T02:24:49.000Z","dependencies_parsed_at":"2023-03-28T00:06:02.235Z","dependency_job_id":null,"html_url":"https://github.com/prouast/ctc-intake-detection","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/prouast%2Fctc-intake-detection","tags_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prouast%2Fctc-intake-detection/tags","releases_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prouast%2Fctc-intake-detection/releases","manifests_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/repositories/prouast%2Fctc-intake-detection/manifests","owner_url":"https://repos.ecosyste.ms/api/v1/hosts/GitHub/owners/prouast","download_url":"https://codeload.github.com/prouast/ctc-intake-detection/tar.gz/refs/heads/master","host":{"name":"GitHub","url":"https://github.com","kind":"github","repositories_count":245878924,"owners_count":20687297,"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":["ctc","deep-learning","intake-gesture-detection","oreba-dataset","tensorflow"],"created_at":"2024-10-15T19:06:14.741Z","updated_at":"2026-05-04T02:35:52.529Z","avatar_url":"https://github.com/prouast.png","language":"Python","funding_links":[],"categories":[],"sub_categories":[],"readme":"# ctc-intake-detection\n\nAutomatic detection of intake gestures with CTC from inertial or video data.\n\n## Data access\n\nFor our experiments, we use the [OREBA](http://www.newcastle.edu.au/oreba) and [Clemson](http://cecas.clemson.edu/~ahoover/cafeteria/) datasets available from the respective websites.\n\n## Data preparation\n\nWe have separate repositories with utilities to generate `TFRecord` files from the raw [inertial data](https://github.com/prouast/inertial-sensor-processing) and [video data](https://github.com/prouast/video-sensor-processing).\n\n## Usage\n\nBuild `tensorflow_ctc_ext_beam_search_decoder` available at [ctc-beam-search-op](https://github.com/prouast/ctc-beam-search-op) for your system and install the wheel via pip, e.g.:\n\n```\npip install tensorflow_ctc_ext_beam_search_decoder-0.1-cp36-cp36m-linux_x86_64.whl\n```\n\nMake sure that all other requirements are fulfilled:\n\n```\n$ pip install -r requirements.txt\n```\n\nThen call `main.py`:\n\n```\n$ python main.py\n```\n\nThe following flags can be set:\n\n| Argument | Description | Default |\n| --- | --- | --- |\n| --batch_size | Training batch size | 128 |\n| --beam_width | Beam width during beam search | 10 |\n| --dataset | Which dataset is used {oreba-dis or clemson} | oreba-dis |\n| --decode_fn | Select the decode_fn {greedy or beam_search} | beam_search |\n| --eval_batch_size | Evaluation batch size | 1 |\n| --eval_dir | Directory with evaluation data | data/inert/valid |\n| --eval_steps | Eval and save best model after every x steps | 1000 |\n| --input_length | Number of input sequence elements | 128 |\n| --input_mode | Select input mode {inertial or video} | inertial |\n| --label_mode | Select the label mode | label_1 |\n| --log_steps | Log after every x steps | 250 |\n| --loss_mode | Select loss mode {ctc or crossent} | ctc |\n| --lr_base | Base learning rate | 1e-3 |\n| --lr_decay_fn | Select learning rate decay fn {exponential or piecewise_constant} | exponential |\n| --lr_decay_rate | Learning rate decay rate | 0.9 |\n| --mixed_precision | Use mixed precision {True or False} | False |\n| --mode | Select mode {train_and_evaluate or predict} | train_and_evaluate |\n| --model | Select model {video_resnet_cnn_lstm or inert_resnet_cnn_lstm} | inert_resnet_cnn_lstm |\n| --model_ckpt | Model checkpoint for prediction (e.g., model_5000) | None |\n| --model_dir | Output directory for model and training stats | run |\n| --num_shuffle | Size of the shuffle buffer | 50000 |\n| --predict_dir | Output directory for prediction mode | predict |\n| --predict_mode | Select aggregation mode for predictions {video_level_average, video_level_concat, batch_level_voted, batch_level_concat, probs} | batch_level_voted |\n| --seq_shift | Shift when generating sequences | 2 |\n| --train_dir | Directory with training data | data/inert/train |\n| --train_epochs | Number of train epochs | 60 |\n| --use_def | Use default class for representation | False |\n","project_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fprouast%2Fctc-intake-detection","html_url":"https://awesome.ecosyste.ms/projects/github.com%2Fprouast%2Fctc-intake-detection","lists_url":"https://awesome.ecosyste.ms/api/v1/projects/github.com%2Fprouast%2Fctc-intake-detection/lists"}