TsetFall dataset

---

A Comprehensive, fine-grained, and publicly available dataset generated for human fall detection.

Table of Contents

• Table of Contents
• About
• Folder structure
• Video descriptions
• Download
• Publications
• Authors

About

Human fall detection research remains an underdeveloped area, particularly with respect to utilizing dynamic images for analysis. Existing fall detection datasets often suffer from size, diversity, and representativeness limitations, as they generally comprise a small number of videos featuring a limited range of actions, camera perspectives, and lighting conditions. Moreover, these datasets typically lack the complexity of real-world fall scenarios, as they often exclude distracting objects and offer only one or two camera angles. To address these challenges, this work introduces TsetFall, a comprehensive, fine-grained, and publicly available dataset generated for fall detection research. Employing a novel AI technique devised by the present authors, the dataset annotation process was notably expedited. By bridging the current gaps in available resources, the TsetFall dataset serves as a valuable benchmark for advancing the field and tackling the multifaceted nature of human fall detection.

Folder structure

The content is made available in the following format:

• tsetfall.rar: A compressed folder containing each of the images extracted from the video sequences. The image name follows the following pattern: cam_number__sequenceNumber-sequenceName__frameNumber.jpeg. Example: cam_1__1-andando-normalmente__00001.jpeg.

• tsetfall_videos.rar: A compressed folder that contains the sequences arranged in video format. The image name follows the following pattern: cam_number__sequenceNumber-sequenceName.mp4.

• ground_truth.csv: The CSV file that contains the annotations performed by humans only. Each line of the file contains the following fields:

  • Image name
  • Class of the objects of interest: Not Fallen (NF), Fallen (FN), Falling (FG), and Confounding (C)
  • 4 fields containing the object's location in x_min, y_min, x_max, y_max format

• ground_truth_extended.csv: The CSV file that contains the AI-assisted annotations. Each line of the file contains the same structure of the previous file, with the addition of the following fields:

  • Detection confidence
  • The algorithm that generated the object detection

tsetfall
├─ tsetfall.rar
│  └─ [list of images]
├─ tsetfall_videos.rar
│  └─ [list of videos]
├─ ground_truth.csv 
└─ ground_truth_extended.csv 

Video descriptions

#	Sequence Name (in Portuguese)	Description
1	andando-normalmente	Walking and interacting with objects.
2	desmaiando	Fainting.
3	deitando-nas-cadeiras	Lying on three chairs.
4	sentando-na-cadeira-com-gatinho	Sitting in a chair with a kitten by the side.
5	sentando-no-chao	Sitting on the floor.
6	andando-com-bengala	Walking with a cane, followed by a fall.
7	andando-com-anda-ja	Walking with a walker, followed by a fall.
8	caindo-de-costas	Falling backwards.
9	caindo-de-barriga	Falling forward.
10	andando-com-casaco-na-cadeira	Walking and interacting with objects, with a coat hanging on a chair.
11	sentando-com-casaco-na-cadeira	Sitting on a chair located next to another chair with a coat hanging on it.
12	cair-no-escuro	Falling forward in the dark.
13	teste-da-camisa-de-botoes	Handling a button-up shirt on a hanger in the dark.
14	polichinelo-no-escuro	Jumping jacks exercise in the dark.
15	apoio-de-frente-escuro	Front support exercise in the dark.
16	abdominal-no-escuro	Abdominal exercise in the dark.
17	flexao-com-pesos-no-escuro	Push-ups with weights exercise in the dark.
18	agachamento-no-escuro	Squats exercise in the dark.
19	busto-do-manequim-no-escuro	Falling backwards while holding a mannequin torso in the dark.
20	entrar-agachar-e-para-no-escuro	Walking to the center of the room, squatting down, standing up, and exiting, all in the dark.
21	andar-agachado-no-escuro	Walking crouched in the dark.
22	andar-com-toalha-turban-no-escuro	Walking with a towel turban in the dark.
23	sentar-e-com-dificuldade	Sit-stand with difficulty in a chair.
24	sentar-levantar-com-dificuldade-no-escuro	Sit-stand with difficulty in a chair in the dark.
25	queda-com-manequim	Falling backwards while holding a mannequin torso.
26	teste-camisa-com-botoes	Handling a button-up shirt on a hanger.
27	polichinelos	Jumping jacks exercise.
28	apoio-de-frente	Front support exercise.
29	abdominal	Abdominal exercise.
30	flexao-com-pesos	Push-ups with weights exercise.
31	agachamento	Squats exercise.
32	entrar-e-agachar	Walking to the center of the room and squatting down.
33	andar-agachado	Walking crouched.
34	toalha-na-cabeca	Walking with a towel turban.
35	sentar-levantar-com-dificuldade	Struggling to walk with hands behind the back, then sitting, standing, and leaving the room.
36	teste-com-cachorro	Walking and interacting with a dog.

Download

To access this content, please complete the following form to acquire the decoding key. You can find the download link here.

Publications

• 2023 IEEE SeGAH

Authors

• @eduardodut
• @mandreia
• @thiagoRCOliveira