Skip to content

Latest commit

 

History

History
49 lines (31 loc) · 2.01 KB

README.md

File metadata and controls

49 lines (31 loc) · 2.01 KB
Raw

Multi Object Tracking Using MobileNet SSD

Implementation of Multi Object Tracking using a pretrained MobileNet SSD with dlib library and OpenCV in Python.

Multi Object Tracking:

Multiple object tracking is the task of tracking more than one object in the video. Here, the algorithm assigns a unique variable to each of the objects that are detected in the video frame. Subsequently, it identifies and tracks all these multiple objects in consecutive/upcoming frames of the video.

SSD MobileNet Architecture:

The SSD architecture is a single convolution network that learns to predict bounding box locations and classify these locations in one pass. Hence, SSD can be trained end-to-end. The SSD network consists of base architecture (MobileNet in this case) followed by several convolution layers:

By using SSD, we only need to take one single shot to detect multiple objects within the image, while regional proposal network (RPN) based approaches such as R-CNN series that need two shots, one for generating region proposals, one for detecting the object of each proposal. Thus, SSD is much faster compared with two-shot RPN-based approaches.

Output

Requirements :

  • dlib
  • opencv-python
  • imutils

Usage :

  • Clone this Repository
git clone https://github.com/ayanbag/Multi_Object_Tracking_with_MobileNetSSD.git
cd Multi_Object_Tracking_with_MobileNetSSD

Then run the following command to install the required dependencies.

pip install -r requirements.txt
  • Now excute the following command :
python multi_object_tracking.py -i <path-to-input>

Note: Our script processes the following command line arguments at runtime:

  • --input or -i : The path to the input video file. We’ll perform multi-object tracking with dlib on this video.
  • --confidence or -c : An optional override for the object detection confidence threshold of 0.2 . This value represents the minimum probability to filter weak detections from the object detector.