Skip to content

Sahil-Chavan/Vehicle_Object_Detection

Repository files navigation

Vehicle_Object_Detection

In the search of my first Computer Vision, Object Detection project, I landed on Vehicle Detection system.

I have built this system using Mobilenet V2 architecture, which based on Single-Shot multibox Detection (SSD) network.

Dataset : I got my dataset through http://cbcl.mit.edu/software-datasets/streetscenes

Model Training : As stated I am using an pre-trained SSD Mobilenet model from Tensorflow Model Zoo, and applied Transfer Learning on the model for 2000 epochs. For this training I used TFOD and COCO object detection api.

End point : One can access the system either by the webpage that I have created using Flask, or by using Postman, by accessing the '/api' endpoint, which receives input image through POST File method and provides output image with predictions in base64 format.

End Result : An system which is capable of recognizing vehicles from an image or from video with great speed due to simplistic architecture of SSD mobilenet and sufficient accuracy.

Future Plans : Next I am going to train in YOLO V4/V5 model for this purpose and compare the results.

Libraries Used : OpenCV, TFOD, COCOapi, Tensorflow 2.X, Pillow, Flask, Imutils, etc.

Shout Out to all these references :

About

No description, website, or topics provided.

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

 
 
 

Contributors

Languages