Add files using upload-large-folder tool

LICENSE

@@ -0,0 +1,5 @@
+This dataset aggregates images and annotations governed by multiple upstream licenses
+(Open Images, COCO, WIDER FACE, Roboflow Universe projects) and Meta SAM 3 terms.
+
+Full terms, citations, and usage conditions are described in README.md.
+This file exists so the Hugging Face Hub `license_link` field can point to a repository file.

README.md

@@ -0,0 +1,129 @@
+---
+license: other
+license_name: multiple-upstream-cc-by-nc-nd-sam
+license_link: LICENSE
+pretty_name: "Person, Face, Package & Home Security Detection (YOLO)"
+tags:
+  - image
+  - computer-vision
+  - object-detection
+  - home-security
+  - yolo
+  - ultralytics
+  - bbox
+  - pytorch
+  - coco
+  - open-images
+  - wider-face
+  - roboflow-universe
+  - sam3
+  - segment-anything
+  - smartpotato
+  - parcel-delivery
+  - person-detection
+  - face-detection
+  - vehicle-detection
+  - pet-detection
+  - autonomous-systems
+annotations_creators:
+  - machine-generated
+language_creators:
+  - found
+task_categories:
+  - object-detection
+task_ids:
+  - face-detection
+  - vehicle-detection
+size_categories:
+  - "100K<n<1M"
+---
+
+# Person, Face, Package & Home Security Detection
+
+A multi-source object detection dataset in **YOLO** format, assembled for training models that recognize **people, faces, parcels**, and categories useful for **home security** (e.g. vehicles, bikes, pets, birds), in the wild. Images come from several public benchmarks and community datasets; **bounding boxes were not taken from the original releases** but were **auto-generated** using a custom pipeline built on Meta’s **Segment Anything Model 3 (SAM 3)**.
+
+---
+
+## What’s inside
+
+| | |
+|---|---|
+| **Task** | Object detection (axis-aligned boxes) |
+| **Format** | YOLO (`images/` + `labels/`, normalized `class cx cy w h`) |
+| **Splits** | `train` / `val` / `test` (see `data.yaml`) |
+| **Classes (9)** | `person`, `face`, `vehicle`, `bike`, `bicycle`, `parcel`, `dog`, `cat`, `bird` |
+
+Configuration for Ultralytics-style layouts lives in `bigsplit/data.yaml` (paths, class index order, and split folders).
+
+---
+
+## Downstream project
+
+This dataset was used to train detectors for **[SmartPotato](https://github.com/OliseNS/SmartPotato)**—see that repository for the application code, configuration, and how these weights are integrated.
+
+---
+
+## Training run & checkpoints (bundled)
+
+An **Ultralytics YOLO** training run and exported weights are included under:
+
+| Path | Contents |
+|------|----------|
+| `smartpotato/run/weights/best.pt` | Best checkpoint from the run (by validation metric) |
+| `smartpotato/run/weights/last.pt` | Final-epoch checkpoint |
+| `smartpotato/run/args.yaml` | Training hyperparameters and run configuration |
+| `smartpotato/run/results.csv` | Per-epoch metrics |
+| `smartpotato/run/*.png`, `*.jpg` | Curves (PR, F1, etc.), confusion matrices, and sample train/val batches |
+
+These artifacts are provided for **reproducibility and convenience**; license obligations from the **dataset** (and from any base weights you used, e.g. COCO-pretrained YOLO) still apply when you redistribute or deploy `*.pt` files.
+
+---
+
+## Data sources
+
+| Source | Role in this dataset |
+|--------|----------------------|
+| [**Open Images**](https://storage.googleapis.com/openimages/web/index.html) | Subset of images used as input to the labeling pipeline |
+| [**COCO**](https://cocodataset.org/) (Common Objects in Context) | Subset of images used as input to the labeling pipeline |
+| [**WIDER FACE**](https://shuoyang1213.me/WIDERFACE/) | Subset of images used as input to the labeling pipeline |
+| [**Roboflow Universe**](https://universe.roboflow.com/) | Multiple public projects/packages; each upstream project carries its own license on Roboflow |
+
+**Annotation method:** Labels were produced automatically with a **custom SAM 3** (Meta) script—i.e. model-assisted segmentation / prompting was converted into YOLO box targets. These annotations are **not** the official Open Images, COCO, or WIDER FACE release labels.
+
+---
+
+## Licensing and attribution (read before use)
+
+This repository is a **composite work**. **You must comply with every license that applies to the images and to any tools used to create the labels.** The maintainer of this dataset is not providing legal advice; when in doubt, review the original terms or seek counsel.
+
+### 1. Upstream image and dataset licenses (by origin)
+
+- **Open Images** — The Open Images project publishes data under terms described on the official site; **individual images retain their own Creative Commons (or other) licenses** as recorded in the dataset metadata. See the [Open Images](https://storage.googleapis.com/openimages/web/index.html) documentation and [openimages/dataset](https://github.com/openimages/dataset) materials for the exact conditions that apply to annotations and redistribution.
+- **COCO** — Use is governed by the [COCO Terms of Use](https://cocodataset.org/#termsofuse). **Images are made available under a Creative Commons Attribution 4.0 International (CC BY 4.0) license** (see the official site for full wording and attribution requirements).
+- **WIDER FACE** — According to the [official WIDER FACE page](https://shuoyang1213.me/WIDERFACE/), the dataset is released under **Creative Commons Attribution–NonCommercial–NoDerivatives 4.0 (CC BY-NC-ND 4.0)**. That means **non-commercial** use and **no redistribution of derivative works** under typical CC BY-NC-ND terms—**verify** against the license text linked from the official site before any reuse or redistribution of portions derived from WIDER FACE.
+- **Roboflow Universe** — Each project lists its own license (often **CC BY 4.0**, but not always). Open the corresponding Universe project page(s) you relied on and follow the stated license and attribution.
+
+### 2. Auto-annotations and Meta SAM 3
+
+Bounding boxes (and any related masks or prompts used in your pipeline) were generated using **SAM 3** from Meta. The SAM 3 codebase and weights are distributed under Meta’s **SAM License** (see [facebookresearch/sam3](https://github.com/facebookresearch/sam3/blob/main/LICENSE)). In short: use and redistribution of SAM Materials are subject to that agreement (including attribution for research outputs, export-control and acceptable-use rules, and redistribution terms). **Keep a copy of the SAM License** with any derivative distribution that is tied to SAM-generated outputs, as required there.
+
+### 3. This dataset’s labels
+
+The **YOLO labels in this repo are a new, automated annotation layer** on top of third-party images. They inherit the **restrictions** of the underlying images **and** the terms governing the tooling (SAM 3) used to create them. **Do not assume** a single SPDX license covers the whole corpus. See the `LICENSE` file for a short pointer; details are in the bullets above.
+
+---
+
+## Citations
+
+If you use this dataset in research, cite the original sources you actually used (Open Images, COCO, WIDER FACE, Roboflow projects) and acknowledge **SAM 3** if your work builds on these auto-annotations. Example references:
+
+- **COCO:** T.-Y. Lin et al., *Microsoft COCO: Common Objects in Context*, ECCV 2014.  
+- **Open Images:** A. Kuznetsova et al., *The Open Images Dataset V4*, IJCV 2020.  
+- **WIDER FACE:** S. Yang et al., *WIDER FACE: A Face Detection Benchmark*, CVPR 2016.  
+- **SAM 3:** See the citation guidance in the [SAM 3 repository](https://github.com/facebookresearch/sam3).
+
+---
+
+## Disclaimer
+
+THE DATASET IS PROVIDED **“AS IS”**, WITHOUT WARRANTY OF ANY KIND. Auto-labels may contain errors, omissions, or biases inherited from base models and source domains. You are responsible for compliance with all applicable licenses, privacy rules, and acceptable-use policies when downloading, training, or deploying models trained on this data.

bigsplit/data.yaml

@@ -0,0 +1,17 @@
+train: images/train          # Training images
+val: images/val              # Validation images
+test: images/test            # Test images
+# Number of classes
+nc: 9
+
+# Class names (exact order: 0-8)
+names:
+  - person
+  - face
+  - vehicle
+  - bike
+  - bicycle
+  - parcel
+  - dog
+  - cat
+  - bird

push_to_hf.sh

@@ -0,0 +1,71 @@
+#!/usr/bin/env bash
+# Upload this dataset directory to the Hugging Face Hub (dataset repo).
+#
+# Prerequisites:
+#   pip install -U huggingface_hub
+#   hf auth login
+#
+# Environment (optional):
+#   HF_USERNAME        Hub username (default: OliseNS)
+#   HF_DATASET_NAME    Repo name only, no org prefix (default: person-face-package-home-security-detection)
+#   HF_USE_LARGE       If 0, uses "hf upload" instead of "hf upload-large-folder" (default: 1)
+#
+# Usage:
+#   ./push_to_hf.sh
+#   HF_USERNAME=myuser ./push_to_hf.sh
+#   ./push_to_hf.sh --sync          # use hf upload (faster for small changes)
+#   ./push_to_hf.sh -- --commit-message "Refresh labels"
+
+set -euo pipefail
+
+SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)"
+ROOT="$SCRIPT_DIR"
+
+HF_USERNAME="${HF_USERNAME:-OliseNS}"
+HF_DATASET_NAME="${HF_DATASET_NAME:-person-face-package-home-security-detection}"
+REPO_ID="${HF_USERNAME}/${HF_DATASET_NAME}"
+HF_USE_LARGE="${HF_USE_LARGE:-1}"
+
+USE_LARGE=1
+EXTRA=()
+while [[ $# -gt 0 ]]; do
+  case "$1" in
+    --sync|--small|--fast)
+      USE_LARGE=0
+      shift
+      ;;
+    --help|-h)
+      sed -n '1,20p' "$0"
+      exit 0
+      ;;
+    --)
+      shift
+      EXTRA=("$@")
+      break
+      ;;
+    *)
+      EXTRA+=("$1")
+      shift
+      ;;
+  esac
+done
+
+if [[ "${HF_USE_LARGE}" == "0" ]]; then
+  USE_LARGE=0
+fi
+
+if ! command -v hf >/dev/null 2>&1; then
+  echo "error: 'hf' not found. Install with: pip install -U huggingface_hub" >&2
+  exit 1
+fi
+
+echo "Repository: datasets/${REPO_ID}"
+echo "Local root: ${ROOT}"
+echo "(First upload creates the dataset repo if it does not exist.)"
+echo
+
+if [[ "$USE_LARGE" -eq 1 ]]; then
+  exec hf upload-large-folder "$REPO_ID" "$ROOT" --repo-type dataset "${EXTRA[@]}"
+else
+  exec hf upload "$REPO_ID" "$ROOT" . --repo-type dataset "${EXTRA[@]}"
+fi

smartpotato/run/args.yaml

@@ -0,0 +1,108 @@
+task: detect
+mode: train
+model: security_camera_project/run_v4_anti_lazy_nano/weights/best.pt
+data: bigsplit/data.yaml
+epochs: 30
+time: null
+patience: 100
+batch: 64
+imgsz: 640
+save: true
+save_period: -1
+cache: false
+device: null
+workers: 8
+project: security_camera_project
+name: run_v4_safe_polish
+exist_ok: false
+pretrained: true
+optimizer: SGD
+verbose: true
+seed: 0
+deterministic: true
+single_cls: false
+rect: false
+cos_lr: false
+close_mosaic: 15
+resume: false
+amp: true
+fraction: 1.0
+profile: false
+freeze: null
+multi_scale: 0.0
+compile: false
+overlap_mask: true
+mask_ratio: 4
+dropout: 0.0
+val: true
+split: val
+save_json: false
+conf: null
+iou: 0.7
+max_det: 300
+half: false
+dnn: false
+plots: true
+source: null
+vid_stride: 1
+stream_buffer: false
+visualize: false
+augment: false
+agnostic_nms: false
+classes: null
+retina_masks: false
+embed: null
+show: false
+save_frames: false
+save_txt: false
+save_conf: false
+save_crop: false
+show_labels: true
+show_conf: true
+show_boxes: true
+line_width: null
+format: torchscript
+keras: false
+optimize: false
+int8: false
+dynamic: false
+simplify: true
+opset: null
+workspace: null
+nms: false
+lr0: 0.0005
+lrf: 0.001
+momentum: 0.95
+weight_decay: 0.0005
+warmup_epochs: 3.0
+warmup_momentum: 0.8
+warmup_bias_lr: 0.1
+box: 10.0
+cls: 4.0
+dfl: 2.0
+pose: 12.0
+kobj: 1.0
+rle: 1.0
+angle: 1.0
+nbs: 64
+hsv_h: 0.015
+hsv_s: 0.7
+hsv_v: 0.4
+degrees: 20.0
+translate: 0.05
+scale: 0.3
+shear: 1.0
+perspective: 0.0
+flipud: 0.0
+fliplr: 0.5
+bgr: 0.0
+mosaic: 0.3
+mixup: 0.0
+cutmix: 0.0
+copy_paste: 0.0
+copy_paste_mode: flip
+auto_augment: randaugment
+erasing: 0.4
+cfg: null
+tracker: botsort.yaml
+save_dir: /train/security_camera_project/run_v4_safe_polish

smartpotato/run/results.csv

@@ -0,0 +1,31 @@
+epoch,time,train/box_loss,train/cls_loss,train/dfl_loss,metrics/precision(B),metrics/recall(B),metrics/mAP50(B),metrics/mAP50-95(B),val/box_loss,val/cls_loss,val/dfl_loss,lr/pg0,lr/pg1,lr/pg2
+1,401.83,2.23567,14.3862,2.07994,0.50743,0.34587,0.35884,0.20206,2.2322,13.8504,1.93987,0.0668539,0.000166563,0.000166563
+2,759.488,2.06565,12.7099,1.96048,0.59814,0.43026,0.47009,0.26847,2.15854,11.9562,1.85863,0.0336762,0.000322133,0.000322133
+3,1142.59,1.89777,11.2328,1.84567,0.69631,0.51594,0.58955,0.37725,1.84403,9.55414,1.66047,0.000487307,0.000466603,0.000466603
+4,1492.56,1.73645,9.85486,1.73947,0.712,0.52866,0.60435,0.38916,1.83455,9.32889,1.64793,0.00045005,0.00045005,0.00045005
+5,1842.99,1.71654,9.69503,1.72718,0.71817,0.52835,0.60817,0.39605,1.79914,9.20603,1.63147,0.0004334,0.0004334,0.0004334
+6,2194.09,1.70788,9.58806,1.71961,0.71429,0.53317,0.61147,0.39858,1.7936,9.15417,1.62669,0.00041675,0.00041675,0.00041675
+7,2584.87,1.69542,9.50983,1.71257,0.72172,0.53251,0.6134,0.40018,1.79022,9.10324,1.62327,0.0004001,0.0004001,0.0004001
+8,2954.25,1.69225,9.46528,1.70889,0.72127,0.53473,0.61569,0.40377,1.77222,9.05541,1.61555,0.00038345,0.00038345,0.00038345
+9,3313.44,1.68604,9.38998,1.70346,0.72163,0.53595,0.61733,0.40501,1.76833,9.02536,1.61254,0.0003668,0.0003668,0.0003668
+10,3677.33,1.67776,9.33915,1.69985,0.72306,0.53694,0.61837,0.40644,1.76353,8.99795,1.61033,0.00035015,0.00035015,0.00035015
+11,4045.97,1.67492,9.30761,1.69819,0.71945,0.53941,0.61926,0.40762,1.75957,8.98113,1.60855,0.0003335,0.0003335,0.0003335
+12,4394.72,1.66786,9.25588,1.69401,0.72248,0.53853,0.61993,0.4083,1.75719,8.96502,1.60729,0.00031685,0.00031685,0.00031685
+13,4745.74,1.66355,9.1992,1.68919,0.72098,0.5394,0.6201,0.40883,1.75443,8.95279,1.60596,0.0003002,0.0003002,0.0003002
+14,5100.54,1.66031,9.16325,1.68855,0.72015,0.54053,0.62096,0.40971,1.75302,8.94084,1.60506,0.00028355,0.00028355,0.00028355
+15,5462.66,1.65625,9.14684,1.68591,0.72049,0.53994,0.62103,0.4102,1.74984,8.92872,1.60355,0.0002669,0.0002669,0.0002669
+16,5823.32,1.60921,8.59339,1.64196,0.72117,0.54047,0.62181,0.41091,1.74702,8.91673,1.60212,0.00025025,0.00025025,0.00025025
+17,6183.37,1.6044,8.5469,1.63885,0.72004,0.5423,0.62245,0.41138,1.74581,8.90695,1.60122,0.0002336,0.0002336,0.0002336
+18,6575.9,1.60247,8.51157,1.63729,0.72106,0.54244,0.623,0.41182,1.74505,8.89587,1.60045,0.00021695,0.00021695,0.00021695
+19,6950.71,1.59779,8.47763,1.63497,0.7207,0.5426,0.62336,0.41238,1.74377,8.88429,1.59959,0.0002003,0.0002003,0.0002003
+20,7336.43,1.59645,8.45805,1.63398,0.72101,0.54348,0.62401,0.41258,1.74402,8.87605,1.59915,0.00018365,0.00018365,0.00018365
+21,7709.74,1.59396,8.43909,1.63169,0.72128,0.54407,0.62436,0.41293,1.74335,8.86717,1.59843,0.000167,0.000167,0.000167
+22,8094.63,1.59151,8.41256,1.63074,0.72176,0.54488,0.62453,0.41318,1.74246,8.85721,1.59789,0.00015035,0.00015035,0.00015035
+23,8466.04,1.58936,8.39994,1.62936,0.72194,0.54515,0.625,0.41355,1.74114,8.84899,1.59693,0.0001337,0.0001337,0.0001337
+24,8835.96,1.58813,8.38,1.62812,0.72273,0.54467,0.62516,0.41362,1.74161,8.84307,1.59662,0.00011705,0.00011705,0.00011705
+25,9221.03,1.5863,8.35972,1.62786,0.72469,0.54363,0.62529,0.41378,1.74081,8.83611,1.59589,0.0001004,0.0001004,0.0001004
+26,9594.09,1.58397,8.34554,1.62458,0.72283,0.54499,0.62526,0.41399,1.74057,8.82926,1.59549,8.375e-05,8.375e-05,8.375e-05
+27,9952.83,1.58342,8.32862,1.62437,0.72258,0.54487,0.62541,0.41437,1.73943,8.82161,1.59483,6.71e-05,6.71e-05,6.71e-05
+28,10312.8,1.58305,8.32494,1.62456,0.72164,0.54608,0.62583,0.41453,1.73851,8.81582,1.59417,5.045e-05,5.045e-05,5.045e-05
+29,10673.1,1.5802,8.317,1.62323,0.72395,0.54524,0.62597,0.41466,1.73895,8.81143,1.59397,3.38e-05,3.38e-05,3.38e-05
+30,11034.7,1.581,8.31174,1.62339,0.72512,0.54514,0.62596,0.41482,1.73791,8.8074,1.59355,1.715e-05,1.715e-05,1.715e-05