Human-Body-Segmentation-Fall
The dataset was reprocessed by the authors based on portions of the Le2i and SisFall datasets. Video clips of ADL and fall actions were converted into images using the OpenCV library, after which a human segmentation neural network was applied to obtain images containing only the human foreground, followed by binarization and morphological erosion and dilation processing. The dataset is divided into two parts, each comprising five categories: Blank (no person), Fall, Like‑fall (unstable balance), Lie, and Stand.
Dataset description and usage context
Dataset Overview
Recommended Datasets
-
Le2i - Laboratoire Electronique, Informatique et Image
- Content: Video data covering four scenes (Home, Coffee room, Office, Lecture room), with annotation files for the Home and Coffee room scenes.
- Format: 320x240, 25 FPS.
- Features: High quality, single‑person, large‑scale dataset.
-
SisFall: A Fall and Movement Dataset
- Content: Video data containing ADL and Fall categories.
- Format: 1440x1080.
- Features: High quality, single‑person, large and diverse dataset.
Other Datasets
- Fall Detection Dataset (FDD)
- Multicam Fall Dataset
- UR Fall Detection Dataset (URFD)
- HIGH QUALITY FALL SIMULATION DATA
- TST FALL DETECTION DATASET V2
Human‑Body‑Segmentation‑Fall Dataset
- Content: Processed from portions of the Le2i and SisFall videos using OpenCV, then a human segmentation neural network to obtain images containing only the human foreground.
- Categories: Blank, Fall, Likefall, Lie, Stand.
- Volume: 253 images for Le2i‑processed and 61 images for SisFall‑processed.
- Download: Google and BaiduDisk links provided; sizes are 29.0 MB and 30.2 MB respectively.
Dataset Table
| Category | Le2i‑processed | SisFall‑processed |
|---|---|---|
| Blank | 7 | 7 |
| Fall | 68 | 15 |
| Likefall | 83 | 13 |
| Lie | 11 | 17 |
| Stand | 84 | 9 |
| Total | 253 | 61 |
Pair the dataset with AI analysis and content workflows.
Once the source passes your review, move straight into summarization, transformation, report drafting, or presentation generation with the JuheAI toolchain.