Swoon Detector

---

This repository is conducted by Lee et al. as a final result of the AI Grand Challenge(Homepage). Our task is to detect a fallen person who needed help. We got 3rd place in this challenge. We created this repository to share our results.

install requirements.txt

pip install -r requirements.txt

execute predict.py

you can test our model by executing python predict.py <your dataset path>.

[input]
<your dataset path> must follow that below file directory structure

/your dataset path
 /video_1
   /video_1_001.jpg
   /video_2_002.jpg
   /...
 /video_2
   /video_2_001.jpg
   /video_2_002.jpg
   /...

you don't have to follow file name, but maintain file structure

[output] After inference, you can get a json file. The structure of the json file is as follows.

{
  'annotations':[
    {
      'file_name': 'video_1_001.jpg', # if two or more swoon people exist in frame
      'box':[{
        'position':[150, 150, 300, 300],   # xyxy
        'confidence_score': '0.9987'
        },
        {
        'position':[560, 560, 900, 900],
        'confidence_score': '0.98'
          }]
    },
    {
      'file_name': 'video_1_002.jpg', # if no swoon person in frame
      'box': []
    },
    {
      'file_name': 'video_1_003.jpg', # if just one swoon person in frame
      'box':[{
        'position': [200, 200, 400, 400] # xyxy
        'confidence_score': '0.899'
        }]
    }
  ]
}

you can make video using generated json file
example

dataset link : https://drive.google.com/drive/folders/1JfEMxKb70GSEEUKMBqr62UFOsMbpPK8s?usp=sharing

TODO

• trainable code