FADE: A Dataset for Detecting Falling Objects around Buildings in Video

Citation Author(s):
Zhigang
Tu
the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China.
Zitao
Gao
the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China.
Zhengbo
Zhang
the State Key Laboratory of Information Engineering in Surveying, Mapping and Remote Sensing, Wuhan University, Wuhan 430079, China.
Chunluan
Zhou
Ant Group co Ltd, Beijing 100020, China
Junsong
Yuan
the Computer Science and Engineering Department, The State University of New York at Buffalo, Buffalo, NY 14260 USA.
Bo
Du
the School of Computer Science, Wuhan University, Wuhan 430072, China.
Submitted by:
Zitao Gao
Last updated:
Sun, 06/30/2024 - 04:49
DOI:
10.21227/dn77-mx42
Data Format:
License:
0
0 ratings - Please login to submit your rating.

Abstract 

Falling objects from buildings can cause severe injuries to pedestrians due to the great impact force they exert. Although surveillance cameras are installed around some buildings, it is challenging for humans to capture such events in surveillance videos due to the small size and fast motion of falling objects, as well as the complex background. Therefore, it is necessary to develop methods to automatically detect falling objects around buildings in surveillance videos. To facilitate the investigation of falling object detection, we propose a large, diverse video dataset called FADE (FAlling Object DEtection around Buildings) for the first time. FADE contains 1,881 videos from 18 scenes, featuring 8 falling object categories, 4 weather conditions, and 4 video resolutions. Additionally, we develop a new object detection method called FADE-Net, which effectively leverages motion information and produces small-sized but high-quality proposals for detecting falling objects around buildings. Importantly, our method is extensively evaluated and analyzed by comparing it with the previous approaches used for generic object detection, video object detection, and moving object detection on the FADE dataset. Experimental results show that the proposed FADE-Net significantly outperforms other methods, providing an effective baseline for future research. The dataset and code are publicly available at https://fadedataset.github.io/FADE.github.io/.

Instructions: 

The annotation file keeps consistent with Pascal VOC and contains two keys: "JPEGImages" and "Annotations". The Instructions on how to utilize the dataset in our baseline model (code: https://fadedataset.github.io/FADE.github.io/) is as follows:

how to train

1. download the dataset.

2. unzip the dataset and put it in the 'dataset' folder.

3. run python train.py to train the model with the default parameters defined in train.py. the usage is as same as generic object detection method's usage. you can also change the parameters by yourself.

how to test

1. create a file named test.txt, and put the path of the test videos in it.

2. run python test.py --val test.txt.

3. the results will be saved in the video folder.

Enviorment

 

CPU: Intel(R) Xeon(R) Gold 5117 CPU @ 2.00GHz GPU: GeForce RTX 3090