Skip to content

videoturingtest/VTT_TRACKING

 
 

Repository files navigation

tracking

caption2

VTT_TRACKING

This repository is for VTT tracking research results. Even if a scene change occurs in the video, it gives the same ID. The configuration is largely based on multi-object tracking using re-id method and the image2text method is a separate module.

We mainly use Image-Text-Embedding method and Person ReID baseline.

current code borrows heavily from Image-Text-Embedding. The images were taken from CUHK PEDES dataset.

Prerequisites

  • NVIDIA GPU + CUDA + CuDNN
  • Matconvnet (Unzip matlab) + Matlab 2017b
  • Pytorch 0.4 + Python 3.6
  • Install requirements

Preprocess Datasets

  • For Visual Tracking
  1. Unzip friends2.zip and after download unzip into /MOT_Re-Id/
  2. Download pre-trained model Download
  3. Locate the pre-trained model into /MOT_Re-Id/model/ft_resNet50/
  • For Image2Text
  1. Download GoogleNews
  2. Download CUHK-PEDES
  3. Pre-trained model (currently uploading in progress)

Useage

  • Visual Tracking

dataset structre:

/MOT_Re-Id/Friends
		└ ep1
		  └ gallery
		    └ 0001 (frame), 0002, 0003, ....
		      └ data
		        └ 0001.png, 0002.png, ... (detection results)

run /MOR_Re-Id/MOT_reid.py

  • Image2Text

run src/find_pic_feature_word2_plus

Output

tracker_results.json has tracking coordinates.

coordinates information is as follows.

"coordinates" : x1, y1, x2, y2, id_number

{
	"dataset": "Friends_EP1",
	"coordinates": [
		[
			252, 338, 584, 819, 1
		],
		[
			688, 376, 951, 748, 2
		],
		[
  ...
}

Todos

  • Write MORE example
  • Currently uploading in progress

Acknowledgements

This work was supported by Institute for Information & communications Technology Promotion(IITP) grant funded by the Korea government(MSIT) (2017-0-01780, The technology development for event recognition/relational reasoning and learning knowledge based system for video understanding)

License

MIT

Releases

No releases published

Packages

No packages published

Languages

  • C++ 48.5%
  • MATLAB 40.8%
  • Python 8.5%
  • C 2.1%
  • M 0.1%