|
Download README.md from Jl-wei/PhoneWatchUI: direct link, hf CLI and curl.
- Browser
- Download file 7.1 kB
-
https://huggingface.co/datasets/Jl-wei/PhoneWatchUI/resolve/main/README.md
- Command line
-
hf download hf://datasets/Jl-wei/PhoneWatchUI/README.md
-
curl -L -o README.md https://huggingface.co/datasets/Jl-wei/PhoneWatchUI/resolve/main/README.md
7.1 kB
| # PhoneWatchUI | |
| This dataset is for research purposes only. | |
| ## Dataset structure | |
| ```text | |
| dataset/ | |
| ├── apple_store/ | |
| │ ├── dataset/ | |
| │ │ ├── apps.jsonl | |
| │ │ ├── images.jsonl | |
| │ │ └── images/<app_id>/{watch,phone}/<position>.png | |
| │ └── script/ | |
| │ ├── app_collect.py | |
| │ ├── watch_app_filter.py | |
| │ ├── image_download.py | |
| │ └── README.md | |
| ├── google_play/ | |
| │ ├── dataset/ | |
| │ │ ├── apps.jsonl | |
| │ │ ├── images.jsonl | |
| │ │ └── images/<app_folder>/<position>.<extension> | |
| │ └── script/ | |
| │ ├── app_collect.py | |
| │ ├── watch_app_filter.py | |
| │ ├── image_download.py | |
| │ ├── device_filter.py | |
| │ ├── requirements.txt | |
| │ └── README.md | |
| ├── README.md | |
| └── requirements.txt | |
| ``` | |
| Each store has its own app records, image records, and image files. `image_path` is relative to that store's `dataset/` directory. Google Play folder names preserve distinctions between case-sensitive app IDs. | |
| ## Fields | |
| JSONL files are UTF-8, with one JSON object per line. Each app has one app record; each gallery position has one image record. Repeated URLs and image bytes are retained. Missing source attributes may be absent or null. | |
| ### `apps.jsonl` | |
| | Field | Description | | |
| | --- | --- | | |
| | `app_id` | Apple numeric App ID or Google Play package ID, stored as a string. | | |
| | `store` | `apple_store` or `google_play`. | | |
| | `title` | App name. | | |
| | `description` | App description. | | |
| | `source_url` | App listing URL. | | |
| | `metadata` | Store-specific app attributes below. | | |
| **Apple App Store:** `metadata.watch_manifest` and `metadata.phone_manifest` each contain `genres`, `screenshots`, and `metadata`. Screenshot entries contain `rank`, `url`, `width`, `height`, and `platform`. The nested `metadata` fields are: | |
| | Fields | Description | | |
| | --- | --- | | |
| | `developer_name`, `developer_id`, `developer_url`, `website_url` | Developer identity and links. | | |
| | `price`, `currency`, `is_free` | Pricing. | | |
| | `rating_average`, `rating_count` | Ratings. | | |
| | `app_version`, `app_version_updated_at` | Version and update date. | | |
| | `compatibility`, `platforms` | Supported platforms; each compatibility entry has `platform` and `requirement`. | | |
| **Google Play:** `metadata` contains the following attributes. BSON-specific values retain MongoDB Extended JSON objects. | |
| | Fields | Description | | |
| | --- | --- | | |
| | `summary` | Short app description. | | |
| | `developer`, `developerId`, `developerInternalID`, `developerWebsite` | Developer identity and website. | | |
| | `genre`, `genreId`, `categories` | App categories. | | |
| | `price`, `currency`, `free` | Pricing. | | |
| | `score`, `ratings` | Rating score and count. | | |
| | `version`, `updated`, `androidVersion`, `androidVersionText`, `androidMaxVersion` | App version, update date, and Android requirements. | | |
| | `screenshots`, `icon`, `headerImage`, `videoImage` | Store media URLs or entries. | | |
| | `permissions`, `datasafety` | Permissions and data-safety disclosures. | | |
| ### `images.jsonl` | |
| | Field | Description | | |
| | --- | --- | | |
| | `image_id` | Position identity: `<store>:<app_id>:<gallery>:<position>`. | | |
| | `app_id`, `store` | Join key to the corresponding app record. | | |
| | `gallery` | Apple: `watch` or `phone`; Google Play: `screenshots`. | | |
| | `position` | Zero-based gallery position. | | |
| | `image_path` | Relative file path; null when no file is available. | | |
| | `width`, `height` | Image dimensions in pixels; null if unavailable. | | |
| | `aspect_ratio` | Width divided by height; null if unavailable. | | |
| | `device` | `watch`, `phone`, or `other`. | | |
| | `device_label_source` | Labeling method and parameters, described below. | | |
| | `source_url` | Source image URL. | | |
| | `sha256` | Recorded SHA-256 of the image bytes; null if unavailable. | | |
| Apple labels follow the source gallery and use `{"method":"explicit_app_store_gallery"}`. Google Play uses `{"method":"aspect_ratio","phone_ratio":1.5}`: square images are `watch`, non-square images with long/short side ratio ≥ 1.5 are `phone`, and remaining or dimensionless images are `other`. These labels identify the source gallery or image shape, not necessarily the devices depicted. | |
| ## Image download | |
| Download images using the `source_url` in each `images.jsonl` record. For Google Play, request the original-resolution URL with `=s0`, as implemented in `image_download.py`. Compute SHA-256 over the downloaded file bytes before resizing or re-encoding, and compare it with the record's `sha256`; a match confirms byte-for-byte identity with the released image. Records with a null hash cannot be verified this way, and changed or unavailable source URLs may prevent retrieval of matching bytes. | |
| ## Script usage | |
| Use Python 3.10+. Run the commands below from the dataset root, with new outputs under `work/`. Keep each store's scripts together. Apple scripts use the standard library; install Google Play dependencies with: | |
| ```sh | |
| python3 -m pip install -r requirements.txt | |
| ``` | |
| ### Apple App Store | |
| Prepare `apple_ids.txt` with one numeric App ID or App Store URL per line. The three steps collect app metadata, select apps with explicit Watch platform support, and download their Watch/iPhone galleries: | |
| ```sh | |
| python3 apple_store/script/app_collect.py --app-ids apple_ids.txt --output work/apple/candidates.jsonl | |
| python3 apple_store/script/watch_app_filter.py --input work/apple/candidates.jsonl \ | |
| --output work/apple/dataset/apps.jsonl | |
| python3 apple_store/script/image_download.py --apps work/apple/dataset/apps.jsonl --output work/apple/dataset | |
| ``` | |
| ### Google Play | |
| Prepare `google_ids.txt` with one package ID per line. The four steps collect app metadata, select watch-related apps, download gallery images, and assign device labels: | |
| ```sh | |
| python3 google_play/script/app_collect.py --ids google_ids.txt --output work/google/candidates.jsonl | |
| python3 google_play/script/watch_app_filter.py --input work/google/candidates.jsonl \ | |
| --output work/google/dataset/apps.jsonl | |
| python3 google_play/script/image_download.py --apps work/google/dataset/apps.jsonl --output work/google/dataset | |
| python3 google_play/script/device_filter.py --input work/google/dataset/images.jsonl \ | |
| --output work/google/dataset/images.jsonl --phone-ratio 1.5 | |
| ``` | |
| `watch_app_filter.py` matches Wear OS, Android Wear, smartwatch, watch face, and watch app/application/version in `title`, `summary`, or `description`. Exact patterns are defined in the script. | |
| Google Play `image_download.py` produces image records without `device` and `device_label_source`; `device_filter.py` adds them. Its default retains every record and supports replacing the input JSONL atomically. To export a subset, use `--device watch`, `--device phone`, or `--device other` with a separate output path; image files are unchanged. | |
| Both stores write `apps.jsonl`, `images.jsonl`, and `images/` in the download output directory. Both app collectors also accept app JSONL as the ID input. Use `--help` on any script for its remaining options. | |